Milankovitch Climate Reinforcements

Global and Planetary Change 40 (2004) 27–48 www.elsevier.com/locate/gloplacha

Milankovitch climate reinforcements

George Kukla*, Joyce Gavin

Lamont Doherty Earth Observatory, Palisades, NY 10964, USA

Received 23 July 2002; accepted 7 May 2003

Abstract

More than a century ago, British scientist John Tyndall argued that increased heating of the oceans was needed to start a glaciation (Tyndall, J., 1872. The forms of water in clouds and rivers ice and glaciers. International Scientific Series. The Werner Company, Akron, OH, 196 pp.). We show that he was essentially right and that the principal cause of Quaternary glaciations was the intensification of the hydrologic cycle by the warming of tropical oceans and increase of equator-to-pole temperature gradient, which led to the growth of land-based ice in the high latitudes. The change was due to decreased obliquity and to the increased intensity of the solar beam in boreal winter and spring at the expense of summer and autumn. This resulted in higher frequency of El Ninõ compared to La Ninã anomalies. Decreased water vapor greenhouse forcing and increased reflection from expanding snowfields were also instrumental in the transition from the last interglacial into the glacial. The current orbital changes, although less extreme, are qualitatively similar. Association of recent positive seasonal anomalies of global mean temperature with El Ninõ events suggests that the ongoing global warming may have a significant, orbitally influenced natural component. The warming could continue even without an increase of greenhouse gases. D 2003 Elsevier B.V. All rights reserved.

Keywords: Glaciations; Insolation; Milankovitch; Climate change; Global warming

1. Introduction remain cool enough in summer to sustain the accumulation of snow and ice from the previous winter The frequency of radiometrically dated paleocli- (Milankovitch, 1941; cf. overview in Charlesworth, matic evidence has been clearly shown to be closely 1957). This theory calls for the climate to respond to a linked with variations of the Earth’s circumsolar orbit relatively modest extraterrestrial forcing caused by the (Broecker and van Donk, 1970; Hays et al., 1976; changed seasonal timing of the earth’s distance from Berger, 1978). Since the end of the 19th century, the the sun (precession), with a delay of several millennia. physical cause of the ice ages has been attributed to However, such a lag in climate response is difficult to the decrease of summer insolation to the high northern explain in the face of the more immediate reaction of latitudes. The argument is that temperatures would global temperature to relatively minor radiative per- turbations, such as those caused by volcanic eruptions (Robock and Mao, 1995). * Corresponding author. Tel.: +1-845-365-8421, fax: +1-845- A different theory of the origin of glaciations was 365-8154. proposed by Tyndall, who explained their occurrence E-mail address: [email protected] (G. Kukla). not by the decrease of insolation to northern lands, but

0921-8181/$ - see front matter D 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0921-8181(03)00096-1 28 G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 rather by the increase of insolation to the oceans. This In the following discussion, we focus on the role of would have provided energy for more evaporation and precession in the transition from the last interglacial to the transfer of extra water from the warm oceans to the glacial when conditions resembled the present. the nivation zones in cold subpolar regions. Tyndall (1872, p. 154) noted that: ‘‘So natural was the association of ice and cold that even celebrated men 2. Insolation pairs assumed that all that is needed to produce a great extension of our glaciers is a diminution of the sun’s The annual march of seasons is caused by the temperature. Had they gone through the foregoing obliquity, which today is 23.5j. In the past, it varied reflections and calculations, they would probably between approximately 22j and 24.5j. During times have demanded more heat instead of less for the of low obliquity, more radiation reaches the low production of a ‘glacial epoch’.’’ Both Tyndall and latitudes and less reaches the poles. The insolation later Croll (1890) argued that to form ice on land, gradient between the equator and the poles increases. extra heat has to be supplied to the oceans. According The change affects the seasons and both hemispheres to them, the key mechanism of the Pleistocene climate equally. change is not the temperature decrease, but rather the From a geocentric perspective, the sun moves north water transfer from the oceans to the land. during winter and spring in what we could refer to as In both theories, obliquity and precession are the the Northbound Hemicycle. It moves south during principal orbital variables affecting past climates. Due summer and autumn in the Southbound Hemicycle to precession, the seasonal timing of the closest (Fig. 1). Throughout the paper, if not otherwise noted, (perihelion) and farthest (aphelion) approach of the we refer to the seasons in the boreal and astronomic earth to the sun changes in about a 19- to 23,000-year sense as delimited by the equinoxes and solstices. The cycle. The difference between the distance in perihe- meteorological seasons start about 3 weeks earlier lion and aphelion, the amplitude of the precessional than the astronomic ones. cycle, is modified by eccentricity which refers to the The seasonal progression of insolation is symmet- departure of the earth’s orbit from a circle. It has a rical along the axis of the solstices. At any given time cycle of about 92,000 years. Obliquity is the inclina- in astronomic winter and spring, there is a tion of the earth’s rotational axis away from the line corresponding instant in autumn and summer when normal to the ecliptic and changes on average about the geometric orientation (but not the strength) of the every 41,000 years. solar beam at the top of the atmosphere (TOA) is the

Fig. 1. Schematic geocentric representation of the annual progression of the sun in the Northbound and Southbound Hemicycles. Insolation pairs with the same geometry of the incoming solar beam shown at ‘‘mid-month’’ positions of the angular calendar of Berger (1978). Solid dots indicate the midpoints of each month of the conventional calendar. G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 29

Table 1 these times paired by equal insolation geometry in the Dates for insolation pairs with equal geometry of the incoming solar Southbound and Northbound Hemicycles as insola- beam. Midpoints of the conventional calendar shown on left correspond to the mean monthly climatology tion pairs. The ‘‘mid-month’’ pairs of Berger (1978) are defined by the equal angular distance from the Jan 16 Nov 25 Feb 14 Oct 27 equinoxes and differ from the midpoints of the con- Mar 16 Sept 28 ventional calendar months (cf. Table 1). Apr 15 Aug 29 May 16 Jul 29 Jun 15 Jun 29 3. Delay of the hydrologic cycle behind insolation Jul 16 May 29 Aug 16 Apr 28 Sept 15 Mar 29 Despite the fact that the length of day, the noon Oct 16 Feb 26 solar zenith angle and the progression of the sun Nov 15 Jan 27 between sunrise and sunset are identical in both Dec 16 Dec 27 members of an insolation pair, the climate is quite different. This is due to the following: (1) The seasonal cycle of the principal elements of same at any particular point of the globe. (The solar the hydrologic cycle—sea surface temperature (SST), beam is defined as the radiant energy received at a atmospheric water vapor, snow and ice cover—lags unit area normal to the sun.) For instance, at the spring behind insolation. As a result of the delay, the pro- equinox in March, the noon solar zenith angle and the portion of incoming radiation reaching the surface that length of the day are the same as at the autumn controls the energy budget of the boundary layer equinox. Similarly, for a day in late April, the inso- differs substantially. For example, in late April, the lation geometry, the noon solar zenith angle and the oceans in the high and the middle northern latitudes length of the day anywhere on the earth are the same are cool, the sea ice is extensive, a relatively large as on a corresponding day in late August. We refer to portion of the Northern Hemisphere is covered by

Fig. 2. Annual cycle of the solar beam intensity at the top of the atmosphere (TOA) in W mÀ 2 and as percent departure from the solar constant, for present (A) (open circles) and 116,000 years ago (B) (full circles). S0 equals solar constant. 30 G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 snow and the air is relatively cold and dry. In late temperatures are near their seasonal peaks, the snow August, the length of day and the daily progression of and ice have melted and the surface heat absorption is the sun are the same. However, the air and water high. The greenhouse reinforcement of incoming

Fig. 3. The timing of perihelion at the equinoxes and solstices for 30 millennia before present and the future 10 millennia. Also for the last interglacial and early glacial between 100 and 140,000 years ago. Marine isotope stages (MIS) after Martinson et al. (1987). Month and angular setting of perihelion (in true longitude measured from autumnal equinox) and obliquity in degrees from Berger and Loutre (1994). Details of precession phases explained in text (WP: warming phase; CP: cooling phase; SWP: stable warming phase; SCP: stable cooling phase). Equinoctial seesaw (ESS) in W mÀ 2 showing the difference in the strength of the solar beam. See text for details. Cold climate intervals shown in black and relatively warm climate stippled. Comparison of the precessional setting with the mean frequency of warm and cold ENSO anomalies as modeled by Clement et al. (1999 and personal communication). G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 31 radiation under the relatively humid atmosphere is at be at the vernal equinox (Fig. 3). All our time its annual peak (Kukla and Karl, 1992). references from present are made from 1950 AD. (2) As a result of the seasonally changing earth– There are two extreme positions of the perihelion sun distance, the strength of the solar beam in the two with diametrically different impacts on the insolation members of an insolation pair differs. pairs. They are shown schematically in Fig. 4. (a) Perihelion at the solstices. The difference in the earth–sun distance and in the strength of the solar 4. Seasonal cycle of the solar beam beam is greatest between early summer and early winter. The intensity of the solar beam is equally Little more than 7 centuries ago, the strength of the strong in both members of the insolation pairs. When TOA solar beam for both members of the insolation perihelion is at winter solstice, most of the increase of pairs was the same. This happens only when perihe- insolation intensity occurs between November and lion occurs at a solstice. At all other times, the sun– January compensated by a decrease between May earth distance and the corresponding intensity of the and July. Today, with perihelion in early January, only solar beam in one member of the insolation pair are either greater or smaller than in the other. Very approximately, the higher solar intensity at perihelion is compensated by lower intensity in aphelion and vice versa, so that the sum of the intensity for the pair remains little changed. The solar constant is the rate at which solar radiation is received at the top of the earth’s atmosphere at a surface perpendicular to the incoming beam. It is calculated for the mean sun–earth distance. Today it is approximately 1360–1370 W mÀ 2. However, the earth’s orbit is elliptical and the mean distance from the sun is achieved for a short time only twice a year. For the remainder of the time, the solar beam is either stronger or weaker than the solar constant. Today, perihelion is reached around the 4th of January when the intensity of the solar beam is f 1410 W mÀ 2. Six months later, in aphelion on July 5, the earth is farthest from the sun and its strength is only f 1320 W mÀ 2. Thus, the solar beam has a pronounced seasonal cycle, which today has a range of about 90 W mÀ 2 or 6.6% of the solar constant. The range was 16.6% at the beginning of the last glacial (Fig. 2). The amplitude of the precession is dictated by the eccentricity of the earth’s orbit that is low today but was high at the end of the last interglacial. The last time winter solstice occurred at perihelion Fig. 4. Schematic representation of the two extreme settings of a was 705 years ago. Summer solstice occurred at precessional cycle. (a) With perihelion at winter solstice and (b) perihelion f 11,515 years ago. In the Holocene with perihelion at vernal equinox. In the first case, the distance from altithermal f 6050 years ago, perihelion was reached the sun and corresponding strength of the solar beam in the at the autumnal equinox, and ca. 17,000 years ago insolation pairs are the same; in the second case, they are considerably different. Circles denoting the ‘‘mid-month’’ positions during the last glacial, it took place at the vernal of the earth on the ecliptic (Berger, 1978) with the same insolation equinox. Two thousand years from now, it will be in geometry of the incoming solar beam have the same fill pattern. early February, and 4500 years from now, it will again Progression of precession is indicated by arrows. 32 G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48

2 weeks after the winter solstice, the solar beam is mer and autumn; (2) a negative autumnal mode when strongest in December and January and the intensity the solar beam is stronger in summer and autumn than in insolation pairs differs little. in winter and spring. The modes change about every (b) Perihelion at the equinoxes. In this situation, 10–11 millennia when perihelion occurs at a solstice. the difference in the strength of the solar beam is A positive ESS switch from an autumnal to a vernal largest between early spring and autumn. It accounts mode occurs at winter solstice and a negative ESS for the major impact of the precessional cycle on the switch at summer solstice. transitional seasons. In early summer and winter, the Positive switches occurred 705 y B.P. (1245 AD) solar beam has a value close to the solar constant. and also at the end of the last interglacial 116,250 When perihelion occurs in March, the beam is stron- years ago. Negative ESS switches occurred at summer gest between February and April and weakest between solstice f 11,515 and f 127,300 years ago. At the August and October. peak of the early glacial substage MIS 5d about 111,000 years B.P., ESS was 225 W mÀ 2 and in a positive mode. The intensity of the solar beam was 5. Equinoctial seesaw then about 1480 W mÀ 2 at the vernal equinox and f 1255 W mÀ 2 at the autumn equinox (Fig. 3).At The difference between the strength of the TOA the peak of the interglacial stage MIS 5e 122,000 solar beam at the vernal and autumnal equinox can be years ago, the beam was strongest in autumn and referred to as the equinoctial seesaw (ESS). It is weakest in spring (Fig. 5). expressed in W mÀ 2. There are two modes of ESS One might assume that the precessional cycle (Fig. 5): (1) a positive vernal mode when the solar would have little impact on climate since the earth’s beam is stronger in winter and spring than in the speed on the ecliptic varies with its distance from the corresponding member of the insolation pair in sum- sun. At perihelion it is more rapid than at aphelion so

Fig. 5. Difference in the strength of the solar beam at mid-month dates of Berger (1978) paired by equal geometry of the incoming solar beam. For the interval from 140,000 to 103,000 years B.P. that includes the last interglacial marine isotope stage (MIS 5e) and the early glacial (MIS 5d). Also shown for the interval from 24,000 years B.P. to 12,000 years in the future that includes the Late Glacial and the Holocene. Insolation from Berger and Loutre (1988) and Berger et al. (1988). G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 33 that the stronger solar beam has a shorter duration. 5.1. Ocean heating However, the portion of short-wave radiation that is absorbed by the climate system and that is reflected The ESS shift toward the vernal mode accelerates depends not only on the angle of the beam but also on the warming of the oceans in the Northern Hemi- the physical properties of the atmosphere and surface. sphere and the tropical oceans in both hemispheres Generally, the local energy budget becomes more (Arkin, 1982). Oceans, the planet’s main receptors of positive when the solar beam is stronger and less solar energy, warming in spring and cooling in au- positive when it is weaker. tumn, follow insolation with a 1- to 2-month delay. The impact of the imbalance of solar forcing in the The seasonal range of SST is relatively large in the transitional seasons is especially strong due to the middle and high latitudes. It is considerably lower in interaction with the seasonal cycles of SST, water the Southern Hemisphere than in the Northern Hemi- vapor and snow cover that is maximized near the sphere (Fig. 6). equinoxes. Correspondingly, we expect the following In the vernal ESS mode, the oceans in Northern responses of climate to the positive (vernal) mode of Hemisphere receive a higher portion of the annual ESS when the solar beam in late February, March and insolation total in the season in which the short-wave April is stronger than in late August, September and radiative forcing is most efficient in raising the October. temperature of seawater. This has a particularly strong

Fig. 6. The current monthly TOA insolation for 45jN and 45jS compared with the zonal mean SST averaged for 1950–1992 (Slutz et al., 1985 and updates; SST data acquired through the IRI climate server at http://ingrid.ldeo.columbia.edu). 34 G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 impact in the equatorial Pacific (Kukla and Gavin, of strong warm (El Ninõ) and cold (La Ninã) events 1992). per 500 years for the past 150,000 years using the model of Zebiak and Cane (1987). This is one of 5.2. El Ninõ Southern Oscillation several models that are used to predict the occurrence of current El Ninõs. It includes the main ENSO El Ninõ Southern Oscillation (ENSO) behavior physics and produces self-sustained interannual oscil- appears to be strongly influenced by seasonal changes lations. It has been relatively successful predicting of solar forcing (Clement et al., 1996). SST in the recent El Ninõ events up to a few seasons in advance. tropical Pacific is highly sensitive to the short-wave Because the model input is parameterized along the radiation in the transitional seasons. The modern SST current climate conditions, the results are most appli- seasonal cycle is most pronounced in the eastern cable to past interglacial/glacial transitions during equatorial Pacific where maxima occur in March which the global climate system is likely to have been and April and minima in late summer and early broadly similar to the current one. autumn. During the last interglacial from 127,000 to 116,000 Using Berger’s (1978) past insolation values, years B.P., the same as during the peak Holocene Clement et al. (1999) calculated the average frequency (Tudhope et al., 2001),ElNinõ was relatively infre-

Fig. 7. Modeled mean frequency of strong warm (black) and cold (gray) ENSO anomalies per 500 years between 106,000 and 118,000 years ago (Clement et al., 1999 and updates; Kukla et al., 2002b), and the reconstructed rate of sea level drop per century (large circles) derived from B18O data of Imbrie et al. (1984), Pisias et al. (1984) and Martinson et al. (1987). We assumed that the sea level drop during MIS 5d was 30 m and was directly proportional to the B18O anomaly. Also shown is the ESS difference in TOA for three ‘‘mid-month’’ insolation pairs (cf. Berger, 1978). G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 35

Fig. 8. Zonal mean temperature departures from the 1961 to the 1990 seasonal mean (Jones, 1994 and updates; data acquired through the IRI climate server at http://ingrid.ldeo.columbia.edu). Averaged for El Ninõ (full circle) and La Ninã (square) seasons from 1950 to 1998 as classified by NOAA’s Climate Prediction Center (2002). Latitudes north positive, south negative; latitudes 30 and 60 marked by dotted line. January–March (winter), April–June (spring), July–September (summer) and October–December (autumn). 36 G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 quent with about 50 to 100 events per 500 years and La If the relationship of the anomalies to surface tem- Ninã dominated. This compares well with the 82 perature was the same as today, then the vernal ESS strong warm events computed in the model with would result in a warmer tropical belt, higher global modern solar forcing. The number of modeled El Ninõ mean temperature, higher meridional temperature gra- events was more than 150 in the penultimate glacial dient and cooler high latitudes in autumn (Fig. 8). The MIS 6 (Kukla et al., 2002a). It was also high in the late current global mean surface air temperature as well as early glacial (MIS 5d) when the frequency of La Ninã the troposphere has been shown to be warmer follow- decreased. Even more interesting is the changed dom- ing El Ninõ events (Jones, 1989; Angell, 2000; Tren- inance of modeled La Ninã to El Ninõ at 115,300 y berth et al., 2002 and references therein). The effect is B.P. closely following the positive ESS switch from apparent up to 6 months later. We furthermore show autumnal to vernal mode 116,250 years ago. Most of that the average global surface temperature during the the change of the modeled ENSO frequency took El Ninõ seasons as categorized by NOAA’s Climate about 2–3 millennia between 113,000 and 116,000 Prediction Center (2002) is higher than normal. Lower years B.P. It was accompanied by a doubling of the than normal temperatures characterize La Ninã seasons growth rate of global ice and sea level drop derived (Fig. 9). At the same time, the high latitudes of both from the oxygen isotope deep-sea record of Martinson hemispheres in autumn are cooler than normal, and the et al. (1987) (Fig. 7). We conclude that ESS has a meridional temperature gradient is larger. The opposite major impact on the character of ENSO with the vernal holds for La Ninã conditions (Fig. 8). mode of ESS resulting in the increased frequency of El The principal question with respect to the dynamics Ninõ anomalies and decrease of La Ninã. of ice ages is how the insolation shifts affected precip-

Fig. 9. (A) The departure of seasonal global mean temperature (1950–2000) from Jones (1994, and updates; data acquired from the Climatic Research Unit at http://www.cru.uea.ac.uk). For average El Ninõ and La Ninã seasons as classified by the Climate Prediction Center (2002). Seasons as in Fig. 7. N = number of years in each seasonal average. (B) The same as (A) but with linear trend removed prior to averaging. G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 37 itation. Studies linking precipitation in high latitudes cover shown in El Ninõ years based on the station data with ENSO are limited by lack of reliable measure- (Brown, 1998). ments. Nevertheless, current ENSO has been linked to Because ENSO is an internal mechanism to the higher moisture flux in the West Antarctic sector climate system, it is difficult to separate the cause and (Bromwich et al., 2000). Attempts to quantify the effect relationship between ENSO and global warm- potential relation of El Ninõ to snow cover in the ing. Whether ENSO is affected by the current global Northern Hemisphere have mixed results. Groisman warming is still unknown (Trenberth and Hoar, 1996; et al. (1994) reported a general discontinuous expan- Rajagopalan et al., 1997; Collins, 2000). However, it is sion of snow extent over the Northern Hemisphere in clear that El Ninõ seasonal anomalies are accompanied autumn and winter and less snow cover in spring and by warmer, and La Ninã anomalies by cooler, global summer during El Ninõ events over the satellite period mean temperature, and the meridional surface temper- of record. Station-based data with longer temporal ature gradient between the low and high latitudes is coverage revealed significant differences from satel- high during El Ninõ and low during La Ninã. The past lite-based analyses for North America with less snow ESS shifts toward frequent El Ninõ events are there-

Fig. 10. Seasonal cycle of top of the atmosphere insolation at 45jN and 45jS compared with the mean hemispheric concentration of atmospheric water vapor for 1988–1999 in mm from Randel et al. (1996 and updated through personal communication). 38 G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 fore likely to be accompanied by higher global mean surface and the autumnal mode with a warmer surface. temperature, cooler high latitudes and an increased The impact is expected to be largest over land in the temperature gradient between the tropics and the poles. Northern Hemisphere (Fig. 11).

5.3. Water vapor greenhouse 5.4. Snow

The principal greenhouse gas is atmospheric water Snow cover expands rapidly in autumn and winter vapor. Its concentration changes considerably during a and retreats relatively slowly in spring and summer. year (Randel et al., 1996) and with it the seasonal An example of the annual snow cover cycle for the forcing of the surface energy budget. In late summer central part of the Eurasian continent between 50j and and autumn, the concentration of atmospheric water 70jE is shown in Fig. 12. The average position of the vapor is significantly higher than in winter and spring. southern front of the snowfields at 73jN began The global cycle is dominated by the Northern Hemi- advancing at the end of August, 1972, when the sphere where the average value peaks in July and cloud-free ground level insolation fell to about 240 August and reaches a minimum in February (Fig. 10). WmÀ 2. The maximum mean snow cover extent was Obviously, if short-wave radiation was stronger in reached at the end of December at 41jN. The snow spring and weaker in autumn, the annual mean green- expansion over 32j of latitude was accompanied by house reinforcement of the surface energy budget an insolation drop of 90 W mÀ 2. The snow line would decrease. The vernal mode of the ESS is retreated after February, when the ground insolation therefore expected to be associated with a cooler was approximately 265 W mÀ 2. By the end of April,

Fig. 11. Zonal differences of mean monthly values of precipitable water for insolation pairs for the Northern Hemisphere and globe. Data from NCEP/NCAR reanalysis project (Kalnay et al., 1996) and acquired from http://ingrid.ldeo.columbia.edu. Percent of total land and ocean area for each zone for the hemisphere and globe is also indicated. G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 39

Fig. 12. Mean latitudinal position of Northern Hemisphere snow cover during 1972 between longitudes 50jE and 70jE. Ground-level, cloud- free insolation at the mean snow line position in full line and open circles (left scale) and the mean latitude of the snow line in solid diamonds and dashed line (right scale). Time around the vernal and autumnal equinoxes stippled. Constructed from data of Kukla (1979). the average snowline stood at 53jN where the cloud- 5.5. Sea ice free insolation was about 395 W mÀ 2. The 130 W mÀ 2 insolation increase at the snowline in March and Local insolation plays a minimum role when the April was accompanied by a snow retreat of only 11j sea ice grows in almost complete darkness in early of latitude, whereas the 110 W mÀ 2 insolation de- autumn, controlled principally by the salinity and crease in September and October was accompanied by temperature of the seawater. The disintegration of a snow cover advance of 18j of latitude. The annual pack ice in spring, however, is directly affected by snow cover cycle in the central part of North Amer- insolation. Rapid pack ice dissipation starts along the ican continent is similar (Kukla, 1979). southern margins of the northern pack ice in March There are several reasons for the different relation- and April, affected by stronger insolation that prevents ships between snow and insolation in the transitional the refreezing of leads opened by wind. A stronger seasons. In March and April, when snow cover is still solar beam in boreal spring is therefore expected to fairly extensive, the greenhouse forcing under cold speed-up the opening of the arctic pack ice, lead to dry air is relatively low and the energy loss to space warmer water in summer and fall and to increase due to surface albedo is high. A large snow-covered temperature contrast between the sea and the land in area supports frequent arctic air outbreaks (Leroux, early autumn and winter. This is the time of peak 1993) that can prolong the duration of the snow. In precipitation in the high northern latitudes (Brom- contrast, in September and October, with the same wich, 1995; Serreze and Hurst, 2000). Ewing and insolation geometry, the land is snow-free, the lakes Donn (1956) suggested that an open arctic would lead still open and the greenhouse forcing in humid air is to the growth of northern land ice. Perhaps they were strong. Precipitation supply in the high northern partly right. latitudes is generally high from mid-summer to early Around Antarctica, the leads start to open, on winter. In any case, it seems obvious that the stronger average, when the noon solar zenith angle in October solar beam in spring at the expense of a weaker beam reaches about 55j (Kukla and Gavin, 1982). A weaker in autumn would facilitate earlier establishment and solar beam in austral spring is therefore expected to longer duration of snow covers. delay the opening of the antarctic pack ice. 40 G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48

Contrary to the north, the antarctic snowfall is latitude oceans in both hemispheres as well as in the substantial in both spring and autumn, so that the middle northern latitudes. It should be accompanied impact of a larger sea ice belt on the ice build-up inland by the increased frequency and intensity of the El can be different from the Northern Hemisphere. It is Nin˜o-like conditions in the tropics and, by analogy also possible that more extensive subantarctic sea ice with the current climate, by a higher global mean correlates with an increased supply of relatively warm temperature. At the same time, the land surface in the water from the southern mid-latitudes to the upwelling middle and high latitudes of the northern continents, zone in the equatorial east Pacific that would intensify where it is strongly influenced by the atmospheric El Nin˜o anomalies (Stott, personal communication). greenhouse effect and surface albedo, is expected to become cooler. The gradient between the low and 5.6. Insolation and the thermohaline circulation high latitudes and between the land and ocean would grow larger, and as a result, the water transfer to the One of the largest seasonal differences in the cold high latitudes would likely increase. Deposition salinity and temperature of the upper ocean exists in of snow and ice in nivation areas should be higher. the source area of the thermohaline circulation in the northern North Atlantic. Here, the highest salinity, lowest temperature and the most active deep water 6. Four phases of the precessional cycle production occur in March and April, whereas the opposite extreme is reached in September and October Based on the expected impact of ESS on the (Fig. 13). We are unaware of any studies that have climate system, and especially on the dynamics of analyzed the impact of insolation on the thermohaline the transitional seasons, we can divide the precession- circulation in this area. Thus, we do not know what al cycle into four phases, each about 5000 years long. role, if any, the insolation in the transitional seasons Climate is affected by many more variables than the play in deep water production. precessional setting, but the relatively stable warm phase and stable cold phase do favor interstadial and 5.7. ESS impact summarized stadial conditions, respectively. The orbital impact is further enhanced by the obliquity, with low obliquity The shift of perihelion timing toward late winter supporting ice growth and high obliquity favoring and early spring is expected to result in warmer low relatively warm climate. Centered around the time of

Fig. 13. Annual progression of mean sea surface temperature (Levitus and Boyer, 1994) and salinity at 65.5jN (Levitus et al., 1994) in the North Atlantic between 0j and 20jW. Data acquired through the IRI climate server at http://ingrid.ldeo.columbia.edu. G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 41 perihelion, these four precessional phases are de- grows quickly in the high latitudes resulting in peri- scribed more fully below (cf. Fig. 3). odic glacier, and ice surges into middle latitudes.

6.1. Stable warm phase (SWP) 6.4. Warming phase (WP)

Centered around the perihelion at the equinox in With perihelion centered around summer solstice in boreal autumn, this phase is characterized by stronger June, this phase is characterized by increasing insola- insolation in astronomic summer and autumn opposed tion intensity in boreal spring and summer, opposed by by weaker insolation in winter and spring. The dif- decreasing intensity in autumn and winter. ESS ference in the intensity of the solar beam in the changes rapidly and switches from positive to negative insolation pairs (ESS) is at its peak. Insolation forcing mode. In response to insolation, the frequency of El in boreal summer and autumn is stronger and in Nin˜o decreases and La Nin˜a dominates. Tropical SST winter and spring weaker. The tropical oceans are decreases and the high latitudes warm. The equator-to- relatively cool and the equator-to-pole temperature pole temperature gradient decreases. High obliquity gradient is low. Ice build-up in the high latitudes is reinforces the warming impact of this phase. slow or negative. The global mean temperature at the beginning of the phase may be low, but increases as a result of snow and ice reduction in the high latitudes. 7. Insolation gradient index High obliquity, which further reduces the meridional insolation gradient, strengthens a warm or mild cli- We propose a new astronomic index showing the mate of this phase. difference between the TOA insolation income at the

6.2. Cooling phase (CP)

With perihelion centered around winter solstice in December, this phase is characterized by increasing insolation intensity in astronomic winter and spring, opposed by a decrease in autumn and summer. The difference in the intensity of the solar beam in the insolation pairs in the transitional seasons (ESS) changes rapidly from negative (autumnal) to positive (vernal) mode. In response to insolation, the frequency of El Nin˜o increases and La Nin˜a decreases. Global mean temperature at the beginning of the phase increases but later drops as a result of snow and ice build-up. Low obliquity reinforces the cooling impact of this phase.

6.3. Stable cold phase (SCP)

With perihelion centered around the equinox in boreal spring, this phase is characterized by a stronger solar beam in astronomic winter and spring in contrast to a weaker intensity in summer and autumn. The equinoctial seesaw is positive and high. In response to Fig. 14. Top of the atmosphere insolation at the equator at the vernal equinox minus insolation at 60jN at the autumnal equinox shown in insolation, the frequency and intensity of El Ninõis WmÀ 2 (lower) and in percent difference from today (upper) high, tropical oceans are relatively warm and the abscissa. Data from Berger and Loutre (1988) and Berger et al. meridional temperature gradient is at a peak. Ice (1988). For future 20 and past 150 millennia. 42 G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 equator at vernal equinox and the insolation income at oscillated near the value of a peak interglacial 60jN at the autumnal equinox. The index is expressed (Shackleton et al., 2002), or to a lesser degree, a glacial in W mÀ 2 or as a percent departure from the current (Petit et al., 1999). Such relatively stable intervals value of 222 W mÀ 2. The index reflects the climatic correspond to the times when perihelion occurred near impact of both the changing obliquity that is symmet- the solstices. They are separated by episodes of rapid rical in both hemispheres and the changing precession change. The fastest coolings appear to have taken that affects the equinoctial seesaw (ESS) and has the place with perihelion in late December, January and largest climatic impact on the tropical oceans and on the February, and the fastest warming when it occurred in high latitudes of the Northern Hemisphere (Fig. 14). June and July. (d) The frequency of El Ninõ was relatively high at times of growing polar ice and low during peak 8. Insolation and paleoclimatic records interglacials and interstadials (Kukla et al., 2002a; Stott et al., 2002). Thus, a relatively warm equatorial Paleoclimatic evidence strongly points to the key ocean coexisted with a high volume of global ice and role of insolation in transitional seasons in past a relatively cold equatorial ocean with a low ice climate change. volume. Accurately dated information from the past (a) The glacial and stadial peaks of marine isotope 30,000 years shows frequent El Ninõs toward the end stages of the last 150,000 years, dated by radiometric of the last glacial (Rittenour et al., 2000) and fewer and thickness ratios, marked by relatively low sea during the warmest part of the Holocene (Sandweiss level and high volume of land-based ice, correlate et al., 1996, 2001; Koutavas et al., 2002).The with March and April perihelion and September and intensity and frequency of El Ninõ increased in the October aphelion. The cold peaks of MIS 5d and MIS last few centuries (Rodbell et al., 1999; Tudhope et 4 are also accompanied by obliquity minima. The al., 2001). This is in general agreement with the model interstadials correlate with perihelion in September of Clement et al. (1999). and aphelion in March (Fig. 3). Interglacials are (e) The sea level of the last interglacial was at least marked by high obliquity. The time scale in the last as high as at present from 128,000 to 116,000 years 30,000 years and between 100 and 150,000 years ago ago (Muhs, 2002). This closely matches the duration is based on relatively accurate radiometric dating. The of the autumnal mode of the equinoctial seesaw. remainder is obtained by interpolation of the thickness During that time, the summer insolation progressed of deep-sea sediments or glacier ice (Imbrie et al., from its temporary maximum 127,000 years ago to its 1984; Martinson et al., 1987; Pisias et al., 1984, Kukla minimum at 116,000 years B.P. (Imbrie et al., 1984; and Gavin, 1992; Petit et al., 1999). Martinson et al., 1987). The early glacial stage MIS (b) The highest correlation of insolation with the 5d lasted from 116,000 to 105,000 years B.P. This is ice volume proxy is with the equinoctial insolation to the duration of the vernal mode of ESS. Obliquity the tropics. Berger et al. (1981) analyzed the degree to decreased from its peak of 24.2j at 131 ka B.P. to a which the varying monthly TOA insolation at selected minimum of 22.2j at 112 ka B.P. (Berger, 1978). latitudes of both hemispheres correlated with the (f) While the subpolar ice grew and peaked tem- radiometrically dated oxygen isotope marine record porarily some 111,000 years ago, the oceans in the of Hays et al. (1976). The marine record represents the low latitudes and the vegetation in the northern middle variation of the total volume of global ice. The highest latitudes remained relatively warm until the time of positive correlation is for September in the three peak ice increase (Ruddiman and McIntyre, 1979; latitudinal belts representing the tropics (25jN, 5jN Kukla et al., 2002b; Lehman et al., 2002). and 15jS). The highest negative correlation was (g) Obliquity was high in the last interglacial and in found in March for all tested latitudes (85jN, 65jN, the Holocene, but low in the peak glacial stadials MIS 45jN, 25jN, 5jN, 15jS, 35jS and 75jS). 5d and MIS 4 (about 65,000 years ago). This poses an (c) Changes in past climate proxies do not follow a apparent dilemma. Given the current seasonal cycle of smooth sinusoidal path, but have plateaus that indicate SST, atmospheric transparency, cloudiness and snow several millennia in which the global ice volume and ice, low obliquity would result in high absorption G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 43 and retention of incoming short-wave radiation by the range of global mean temperature of about 3.7 jC planet. More energy would reach the tropical oceans, (Jones et al., 1999) does not differ much from the which have a low albedo and high heat capacity, and modeled difference of 4–5 jC between today and the less would be reflected from the polar caps. One could last glacial 18,000 years ago (Kutzbach and Guetter, expect that in this situation the Earth would heat up, 1984; Rind, 1987). Thus, the principal feature of the but paleoclimatic evidence shows that times of poten- glacial world, the same as the current boreal winter, tially the highest absorption of the sun’s radiation was not so much the lower mean global surface correlate with coldest global climates and largest temperature, but rather the large temperature gradient volume of ice (Kukla, 1969). between the warm oceans and the cold subpolar lands. A similar conclusion can be made for the current It is this high ocean–land temperature contrast that seasonal cycle of global mean temperature (Fig. 15). enabled the transfer of seawater onto land and sup- The highest global surface temperature is reached in ported the growth of ice, which Tyndall (1872) saw as July when the solar radiation reaching the oceans is the main cause of a glaciation. lowest, but the equator-to-pole temperature gradient is minimal. The lowest global mean temperature is attained in January, when the largest amount of 9. Start of the last glacial insolation is channeled to the oceans and when the temperature difference between the equator and the The tentative hypothetical reconstruction of the poles is greatest. Interestingly, the current seasonal last interglacial/glacial transition from currently avail-

Fig. 15. Hemispheric and global monthly mean temperature for 1961–1990 from Jones et al. (1999) plotted against the temperature difference between the equator and the arctic and antarctic circles. The plot for the global mean temperature shows the difference between 5jN–5jS and 60–65jN and S. Data acquired from http://www.cru.uea.ac.uk. 44 G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 able paleoclimatic data (Kukla et al., 2002b) is as conventional one. Instead of viewing the increased follows: albedo and the related drop of land temperature as the only trigger of global climate change, we see (a) The dominant strength of the solar beam in boreal the oceans in the low latitudes as the key recipient autumn magnified by high obliquity characterized of the insolation signal. The warming of low the quasi-stable last interglacial climate that was latitude oceans and the increased gradient between broadly similar to the current one. the warmer oceans and cooler land were combined (b) The ESS shift from autumnal to vernal mode with decreased water vapor greenhouse forcing in 116,000 years B.P. was closely accompanied by autumn and with the earlier establishment of the increase of the El Ninõ frequency and intensity snowfields. This led to the accumulation of polar and decrease of La Ninã. Tropical oceans warmed. ice and drop of sea level. The relative impact of Simultaneously, the northern lands cooled due to orbital variations, as opposed to the dynamics of the decreased greenhouse forcing and earlier ice and ocean currents, probably decreased during development of seasonal snowfields. Water bodies the progression of an interglacial/glacial cycle. that warmed earlier in spring and remained warm longer into the autumn provided additional moisture for the build-up of high latitude ice. 10. Future outlook Weaker insolation in autumn intensified the temperature difference between relatively warm The current orbital configuration, in qualitative waters and cooler lands. terms, is similar to that of 115,540 years ago at the (c) The meridional circulation vigor in boreal autumn start of early glacial stage MIS 5d. At that time, the and winter, driven by increased insolation and same as today, obliquity was decreasing from its early temperature gradients between colder high lati- interglacial peak to its early glacial minimum. How- tudes and warmer tropics, strengthened. The high ever, today the ESS amplitude is only 40% of that at temperature contrast between the oceans and the the end of the last interglacial. It is more similar to the land enabled the intensified transfer of ocean water one some 400,000 years ago, in the first part of the onto the land-based ice. longer lasting MIS 11 or Holstein interglacial (Berger (d) Increased accumulation of snow in nivation zones and Loutre, 2002; Kukla, 2003). The outlook for the led to the growth of glaciers in high latitudes. The natural component of the future climate change is changed distribution of the ice mass accelerated therefore a long continuation of relatively benign the outflow of ice into the forelands and open climates. The artificial increase of greenhouse gases ocean. Sea level dropped. to levels higher than at any time in the last half million (e) Episodic surges from glacier margins into the years complicates the picture, since both natural as ocean lowered SST and salinity, which in turn well as artificial impacts on climate are expected to extended the duration of seasonal pack ice and lead to global warming. Global warming by itself is enabled more intense and farther reaching arctic therefore not a sign of the anthropogenic impact, air outbreaks (Leroux, 1993). The thermohaline especially when it is dominated by the low latitudes. circulation was affected (Broecker, 1991). This A more serious question is whether the solar and the may have led to alteration of oceanic circulation CO2-related global warming would result in the op- worldwide and contributed to the flip-flop behav- posite change of ice volume and sea level. In this ior of local climates (Bond and Lotti, 1995). respect, analysis of principal general circulation mod- (f) Ultimately, a major outbreak of polar ice into the els by Ye and Mather (1997) that shows an increased oceans cooled the subtropical and tropical oceans to snowfall rate in the high latitudes under a doubling of such a degree that the meridional temperature and CO2 needs to be considered. In either case, the precipitation gradients decreased (Bush and Philan- probable development of future climate requires anal- der, 1998; Broecker, 1991) and the glaciers starved. ysis of both components of the ongoing climate (g) Our understanding of the mechanism of Pleisto- change, the one principally related to the orbital cene climate change differs substantially from the modulation of the short-wave radiation income and G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 45 the one related to the long-wave radiation exchanges. 90j away from the equinoxes. Although the true This will require an intense modeling effort. month in the angular calendar ranges between 28 and 34 days, average length of 30.5 days can be used without producing a serious bias (Joussaume and 11. Modeling past climates Braconnot, 1997). The insolation geometry at any day of this angular calendar is the same in the opposite Modeling of past climates is complicated by the member of the insolation pair, and the difference in changing length of the astronomic seasons (Jous- solar intensity in transitional seasons is correctly saume and Braconnot, 1997). During the last million reflected. However, this difference disappears when years, the length of any individual season varied the insolation income is averaged over the varying between 82 and 100 days (Berger and Loutre, length of the interval. Then, even the models using 1994). However, the TOA energy received during an angular subdivision may reflect only the impact of the astronomic season of variable length is constant, solstices (Gallimore and Kutzbach, 1995). irrespective of precessional timing. If a month is considered to be approximately one third of a season, then the TOA insolation income, delivered by a 12. Conclusion stronger solar beam, is compensated by its shorter duration. We propose that the main impact of past orbital Our conventional calendar is set to perihelion changes on climate was not in the varying summer occurring at winter solstice. The length of astronomic insolation to the high latitudes, but rather in the spring and summer is 93 days each, and that of changed strength of the solar beam in early spring autumn and winter 89 days each. The strength of and autumn. At the last interglacial/glacial transition, the solar beam in both members of the insolation pairs this shift led to a warming of low latitude oceans and is approximately the same. Monthly mean climate cooling of the northern lands. The increased equator- parameters and other physical relationships between to-pole and ocean-to-land temperature gradients facil- insolation and climate derived from recent climatolo- itated the poleward transfer of water onto land-based gy may not be justified and applicable to past con- ice. The earlier ocean warming, combined with de- ditions. In particular, this is the case for the vernal and creased water vapor greenhouse forcing over land in autumnal ESS modes. spring and earlier establishment of snowfields in In order to reflect the impact of ESS, a separate autumn, led to the growth of ice sheets and to calendar for each particular timing of the perihelion intermittent episodes of accelerated calving. Our hy- would be required. Otherwise, the equinoxes and pothetical explanation focuses on the first f 20,000 solstices would fall at widely different calendar days years of a glacial cycle. It does not address the further and the use of the monthly mean climate parameters progression of a glaciation and the processes that would produce serious biases (Joussaume and Bra- caused the collapse of the ice sheets some 100,000 connot, 1997). Model calculations would have to years later. We have mentioned only briefly the role of proceed in daily steps based on physical relationships obliquity, which is particularly important in the high completely independent of the parameterizations of latitudes. current climate to avoid this bias. To our knowledge, Our hypothesis is based on correlation of radio- this has not been done. Thus, with the exception of metrically dated paleoclimatic evidence with past Clement et al.’s (1999) ENSO study, the ESS impact orbital variations computed from the laws of celestial on past climates has not yet been properly quantified. mechanics. It is not yet supported by any climate Aware of the problem, Berger (1978) calculated the model other than that of Clement et al. (1999). Even so-called ‘‘mid-month’’ insolation values for 12 angu- with this limitation, the existing data demonstrate that lar positions, 30j apart from the vernal equinox on the the past orbital shift, qualitatively analogous, but 360j annual orbit of the Earth. The date of the vernal stronger than the ongoing one, was accompanied by equinox is arbitrarily taken as March 21st and referred the warming of tropical oceans, the probable increase to as ‘‘mid-March’’. The summer solstice is always of global mean temperature and by the growth of 46 G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 polar ice. Thus, the current global warming may be a umes and the O18 record in deep-sea cores. Reviews of Geo- product of both man-made and natural causes. physics and Space Physics 8, 169–181. Bromwich, D., 1995. Ice sheets and sea level. Nature 373, 18–19. Bromwich, D.H., Roger, A.N., Kallberg, P., Cullather, R.I., White, J.W.C., Kreutz, K.J., 2000. ECMWF analysis and reanalysis Acknowledgements depiction of ENSO signal in Antarctic precipitation. Journal of Climate 13, 1406–1420. We thank Antonio Brambati and Tullio Quaia for Brown, R., 1998. El Ninõ and North American snow cover. Pro- encouragement and patience during the completion of ceedings 55th Eastern Snow Conference, Jackson, New Hamp- this work. We are grateful to Amy Clement, Marie- shire. CRYSYS, Canada, pp. 165–172. Bush, A.B.G., Philander, S.G.H., 1998. The role of ocean–atmos- France Loutre and Andre Berger for use of their phere interactions in tropical cooling during the last glacial unpublished data and for helpful discussions and to maximum. Science 279, 1341–1344. Denis Rousseau for valuable suggestions. We thank Charlesworth, J.K., 1957. The Quaternary Era St. Martin’s Press, Ulrich Mu¨ller, Alexander Prokopenko and other New York. 1700 pp. reviewers for constructive comments that resulted in Clement, A.C., Seager, R., Cane, M.A., Zebiak, S.E., 1996. An ocean dynamical thermostat. Journal of Climate 9, 2190–2196. an improved paper. Partial support was provided by Clement, A., Seager, R., Cane, M., 1999. Orbital controls on the El grant ATM-01-00830 of the Paleoclimate Program of Ninõ/Southern Oscillation tropical climate. Paleoceanography the National Science Foundation. This is a Lamont- 14, 441–456. Doherty Earth Observatory contribution 6396. Climate Prediction Center, 2002. http://www.cpc.ncep.noaa.gov/ products/analysis_monitoring/ensostuff/ensoyears.html (as of 1/1/02). Collins, M., 2000. The El-Ninõ Southern Oscillation in the second References Hadley Centre coupled model and its response to greenhouse warming. Journal of Climate 13, 1299–1312. Angell, J.K., 2000. Tropospheric temperature variations adjusted Croll, J., 1890. Climate and Time Edward Stanford, London. 577 pp. for El Ninõ, 1958–1998. Journal of Geophysical Research Ewing, M., Donn, W., 1956. A theory of ice ages. Science 123, 105 (D9), 11841–11849. 1061–1066. Arkin, A., 1982. The relationship between interannual variability in Gallimore, R.G., Kutzbach, J.E., 1995. Snow cover and sea ice the 200 mb tropical wind field and the Southern Oscillation. sensitivity to generic changes in Earth orbital parameters. Jour- Monthly Weather Review 110, 1393–1404. nal of Geophysical Research 100 (D1), 1103–1120. Berger, A., 1978. Long-term variations of daily insolation and Qua- Groisman, P.Y., Karl, T.R., Knight, R.W., Stenchikov, G.L., ternary climate changes. Journal of the Atmospheric Sciences 1994. Changes of snow cover, temperature, and radiative heat 35, 2362–2367. balance over the Northern Hemisphere. Journal of Climate 7, Berger, A., Loutre, M.F., 1988. Une nouvell solution astronomique 1633–1656. pour les 10 derniers millions d’annees. Scientific Report 1988/ Hays, J.D., Imbrie, J., Shackleton, N.J., 1976. Variations in the 14. Institut d’Astronomie et de Geophysique G. Lamaitre Uni- Earth’s orbit: pacemaker of the ice ages. Science 194, versite Catholique de Louvain, Louvain-la-Neuve. 1121–1132. Berger, A., Loutre, M.F., 1994. Precession, eccentricity, obliquity, Imbrie, J., Hays, J.D., Martinson, D.G., McIntyre, A., Mix, A.C., insolation and paleoclimates. NATO ASI Ser. Long-term Cli- Morley, J.J., Pisias, N.G., Prell, W.L., Shackleton, N.J., 1984. matic Variations. Springer, Berlin, pp. 107–145. The orbital theory of Pleistocene climate: support from a Berger, A., Loutre, M.F., 2002. An exceptionally long interglacial revised chronology of the marine delta 18O record. In: Berger, ahead? Science 297, 1287–1288. A., Imbrie, J., Hays, J., Kukla, G., Saltzman, B. (Eds.), Mi- Berger, A., Guiot, J., Kukla, G., Pestiaux, P., 1981. Long-term lankovitch and Climate, Part I. Reidel Publishing Co., Boston, variations of monthly insolation as related to climatic changes. pp. 269–305. Geologischen Rundschau 70 (2), 748–758. Jones, P.D., 1989. The influence of ENSO on global temperatures. Berger, A., Loutre, M.F., Lasker, J., 1988. Insolation values for the Climate Monitor 17, 80–89. climate of the last 10 million years. Scientific Report 1988/13. Jones, P.D., 1994. Hemispheric surface air temperature variations: Institut d’Astronomie et de Geophysique G. Lamaitre Universite a reanalysis and an update to 1993. Journal of Climate 7, Catholique de Louvain, Louvain-la-Neuve. 1794–1802. Bond, G.C., Lotti, R., 1995. Iceberg discharges into North Atlantic Jones, P.D., New, M., Parker, D.E., Martin, S., Rigor, I.G., 1999. on millennial time scales during last glaciation. Science 267, Surface air temperature and its changes over the past 150 years. 1005–1010. Reviews of Geophysics 37, 173–199. Broecker, W.S., 1991. The great ocean conveyor. Oceanography 4 Joussaume, S., Braconnot, P., 1997. Sensitivity of paleoclimate (2), 79–89. simulation results to season definitions. Journal of Geophysical Broecker, W.S., van Donk, J., 1970. Insolation changes, ice vol- Research 102 (D2), 1943–1956. G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48 47

Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Levitus, S., Burgett, R., Boyer, T.P., 1994. Salinity. World Ocean Gandin, L., Iredell, M., Saha, S., White, G., Woollen, J., Zhu, Atlas, vol. 3. NOAA, NESDIS, US Department of Commerce, Y., Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, Washington, DC. 99 pp. K.C., Ropelewski, C., Wang, J., Leetmaa, A., Reynolds, R., Martinson, D.G., Pisias, N.G., Hays, J.D., Imbrie, J., Moore, T.C.J., Jenne, R., Joseph, D., 1996. NCEP/NCAR 40-year reanalysis Shackleton, N.J., 1987. Age dating and orbital theory of ice project. Bulletin of the American Meteorological Society 77, ages: development of a high-resolution 0 to 300,000-year chro- 437–471. nostratigraphy. Quaternary Research 27, 1–29. Koutavas, A., Lynch-Stieglitz, J., Marchitto Jr., T.M., Sachs, J.P. Milankovitch, M., 1941. Kanon der Erdbestrahlung und seine 2002. El Ninõ-like pattern in ice age tropical Pacific sea surface Anwendung auf das Eiszeitenproblem. Academie Royale temperature. Science 297, 226–230. Serbe, Edition Spećiales, Tome CXXXIII, Section des Scien- Kukla, G., 1969. The cause of Holocene climate change. Geologie ces Mathe´matiques et Naturelles, Tome 33. Stamparija Mihaila en Mijnbouw 48, 307–334. Curcica-Beograd. Kukla, G., 1979. Recent changes in snow and ice. In: Gribbin, J. Muhs, D.R., 2002. The duration of the last interglacial period: (Ed.), Climatic Change. Cambridge Univ. Press, Cambridge, evidence from TIMS uranium-series ages of corals from tectoni- pp. 114–129. cally stable coastlines. Quaternary Research 58, 36–40. Kukla, G., 2003. Continental records of MIS 11. In: Droxler, A.W., Petit, J.R., Jouzel, J., Raynaud, D., Barkov, N.I., Barnola, J.-M., Pore, R.Z., Burckle, L.H. (Eds.), Earth’s Climate and Orbital Basile, I., Bender, M., Chappellaz, J., Davis, M., Delaygue, G., Eccentricity. The Marine Isotope 11 Question. American Geo- Delmotte, M., Kotlyakov, V.M., Legrand, M., Lipenkov, V.Y., physical Union, pp. 207–211. Lorius, C., Pepin, L., Ritz, C., Saltzman, E., Stievenard, M., Kukla, G., Gavin, J., 1982. Potential impact of snow and ice var- 1999. Climate and atmospheric history of the past 420,000 years iations on the heat budget. Proceedings of the Sixth Annual from the Vostok ice core, Antarctica. Nature 399, 429–436. Climate Diagnostics Workshop. US Department of Commerce, Pisias, N.G., Martinson, D.G., Moore, T.C.J., Shackleton, N.J., NOAA, Washington, DC, pp. 64–71. Prell, W., Hays, J., Boden, G., 1984. High resolution strati- Kukla, G., Gavin, J., 1992. Insolation regime of the warm to cold graphic correlation of benthic oxygen isotopic records spanning transitions. In: Kukla, G., Went, E. (Eds.), Start of a Glacial. the last 300,000 years. Marine Geology 56, 119–136. NATO ASI Series I: Global Environmental Change. Springer- Rajagopalan, B., Lall, U., Cane, M.A., 1997. Anomalous ENSO Verlag, Berlin, pp. 307–339. occurrences: an alternate view. Journal of Climate 10, Kukla, G., Karl, T., 1992. Seasonal cycle of insolation and climate 2351–2357. change. Proceedings of the Sixteenth Annual Climate Diagnos- Randel, D.L., Vonder Haar, T.H., Ringerud, M.A., Stephens, G.L., tics Workshop. US Dept. of Commerce, NOAA, pp. 215–222. Greenwald, T.J., Combs, C.L., 1996. A new global water vapor Kukla, G.J., Clement, A.C., Cane, M.A., Gavin, J.E., Zebiak, S.E., dataset. Bulletin of the American Meteorological Society 77, 2002a. Last interglacial and early glacial ENSO. Quaternary 1233–1246. Research 58, 27–31. Rind, D., 1987. Components of the ice age circulation. Journal of Kukla, G.J., Bender, M.L., de Beaulieu, J.-L., Bond, G., Broecker, Geophysical Research 92 (D4), 4241–4281. W.S., Cleveringa, P., Gavin, J.E., Herbert, T.D., Imbrie, J., Rittenour, T.M., Brigham-Grette, J., Mann, M.E., 2000. El Ninõ- Jouzel, J., Keigwin, L.D., Knudsen, K.-L., McManus, J.F., like climate teleconnections in New England during the Late Merkt, J., Muhs, D.R., Muller, H., Poore, R.Z., Porter, S.C., Pleistocene. Science 288, 1039–1042. Seret, G., Shackleton, N.J., Turner, C., Tzedakis, P.C., Wino- Robock, A., Mao, J., 1995. Volcanic signal in surface temperature grad, I.J., 2002b. Last interglacial climates. Quaternary Re- observations. Journal of Climate 8, 1086–1103. search 58, 2–13. Rodbell, D.T., Seltzer, G.O., Anderson, D.M., Abbott, M.B., En- Kutzbach, J.E., Guetter, P.J., 1984. Sensitivity of Late-glacial and field, D.B., Newman, J.H., 1999. An f 15,000-year record of Holocene climates to the combined effects of orbital parameter El Ninõ-driven alluviation in southwestern Ecuador. Science changes and lower boundary condition changes: ‘‘snapshot’’ 283, 516–520. simulations with a general circulation model for 18, 9 and 6 Ruddiman, W.F., McIntyre, A., 1979. Warmth of the subpolar North ka BP. Annals of Glaciology 5, 85–87. Atlantic Ocean during Northern Hemisphere ice-sheet growth. Lehman, S.J., Sachs, J.P., Crotwell, A.M., Keigwin, L.D., Boyle, Science 204, 173–175. E.A., 2002. Relation of subtropical Atlantic temperature, high Sandweiss, D.H., Richardson, J.B., Reitz, E.J., Rollins, H.B., latitude ice rafting, deep water formation, and European climate Maasch, K.A., 1996. Geoarchaeological evidence from Peru for 130,000–60,000 years ago. Quaternary Science Reviews 21, a 5000 years B.P. onset of El Ninõ. Science 273, 1531–1533. 1917–1924. Sandweiss, D.H., Maasch, K.A., Berger, R.L., Richardson, J.B., Leroux, M., 1993. The mobile polar high: a new concept explaining Rollins, H.B., Clement, A., 2001. Variation in Holocene El Ninõ present mechanisms of meridional air-mass and energy ex- frequencies: climate records and cultural consequences in an- changes and global propagation of palaeoclimatic changes. cient Peru. Geology 29, 603–606. Global and Planetary Change 7, 69–93. Serreze, M.C., Hurst, C.M., 2000. Representation of mean Arctic Levitus, S., Boyer, T., 1994. Temperature. World Ocean Atlas, vol. precipitation from NCEP–NCAR and ERA reanalysis. Journal 4. NOAA, NESDIS, US Department of Commerce, Washington, of Climate 13, 182–201. DC. 117 pp. Shackleton, N.J., Chapman, M., Sanchez-Goni, M.F., Pailler, D. 48 G. Kukla, J. Gavin / Global and Planetary Change 40 (2004) 27–48

Lancelot, Y., 2002. The classic marine isotope substage 5e. pheric surface temperatures. Journal of Geophysical Research Quaternary Research 58, 14–16. 107 (D8), AAC5-1–AAC5-19. Slutz, R.J., Lubker, S.J., Hiscox, J.D., Woodruff, S.D., Jenne, Tudhope, A.W., Chilcott, C.P., McCulloch, M.T., Cook, E.R., R.L., Steurer, P.M., Elms, J.D., 1985. Comprehensive Ocean- Chappell, J., Ellam, R.M., Lea, D.W., Lough, J.M., Shim- Atmosphere Data Set; Release 1. Climate Research Program, mield, G.B., 2001. Variability in the El Ninõ Southern Oscil- Boulder, CO. lation through a glacial – interglacial cycle. Science 291, Stott, L., Poulsen, C., Lund, S., Thunell, R., 2002. Super ENSO and 1511–1517. global climate oscillations at millennial time scales. Science Tyndall, J., 1872. The forms of water in clouds and rivers ice and 297, 222–226. glaciersInternational Scientific Series. The Werner Company, Trenberth, K.E., Hoar, T.J., 1996. The 1990–1995 El Ninõ-South- Akron, OH. 196 pp. ern Oscillation event: longest on record. Geophysical Research Ye, H., Mather, J.R., 1997. Polar snow cover changes and global Letters 23 (1), 57–60. warming. International Journal of Climatology 17, 155–162. Trenberth, K.E., Caron, J.M., Stepaniak, D.P., Worley, S., 2002. Zebiak, S.E., Cane, M.A., 1987. A model El Ninõ-Southern Oscil- Evolution of El Ninõ-Southern Oscillation and global atmos- lation. Monthly Weather Review 115, 2262–2278.