Climate proxy records of very long duration (e.g., ice cores,
sea sediment cores) allow discrimination of climate variability
on millennial timescales. Hammer [1980] used conductivity
measurements of ice cores to record the dates and frequency of
massive volcanic eruption in the Northern Hemisphere. His
analysis highlighted a frequency of approximately one major
eruption every thousand years from 1390 B.C. (Thera, Aegean Sea)
to 1912 A.D. (Katmai, Alaska). Because volcanic eruptions often
release substantial quantities of aerosols to the atmosphere,
their occurrence influences albedo, cloud formation and thus,
climate. The 40,000-year record of dust concentrations measured
in the ice of the Dunde core, China [ Thompson et al., 1989]
reflected several-thousand year periods of warming and cooling,
in which more abrupt changes were embedded. The 41,000-year
GISP2 record from Greenland showed atmospheric response to
massive iceberg discharge events, with major ion, dust, and
isotope peaks at frequencies of a few to several thousand years
[ Mayewski et al., 1994]. The 250,000-year Greenland record
studied by Dansgaard et al. [1993] also permits examination
of the frequency and timing of glacial-interglacial cycling. The
core revealed 24 interstadials (IS #1-24) from the Bø lling
(
14 kyr BP) to 
110 kyr BP, which range from nearly
10,000 years apart to 
2,000 years apart (e.g., IS #16, 17).
Jouzel et al. [1993] examined the Antarctic Vostok core and
found rates of temperature change as high as 
12
C over
15 kyr. Yiou et al. [1991] used nonparametric spectral
techniques to investigate cyclic patterns of climate variation in
ice core records. Their analyses revealed stable cycles with
periods of 11.1, 6.0, 4.4, 3.5, 2.7, and 2.4 kyr. The irregular
pattern of cyclicity led the authors to conclude that nonlinear
feedback mechanisms must play an important role in paleoclimatic
variability.
Sea sediments, with the exception of cores from the Norwegian
Sea [ Lehman and Keigwin, 1992], are often not of high
enough resolution to record fast oscillations of the climate
system, but they are of long enough duration to record rates of
change for slower events. Ruddiman and McIntyre [1981]
studied carbonate productivity in sea sediment cores from the
North Atlantic to determine rates of change of ice volume during
the last deglaciation. They concluded that slightly more than
50% of Northern Hemisphere ice disappeared in 3,000 yrs (16 to 13
kyr BP). Despite the oceanic cooling of 7
to 10
C
over the 1,000-year Younger Dryas event (
12-11 kyr BP) the
present sea ice regime was fully established by 6,000 yr BP [
Ruddiman and McIntyre, 1981]. Recently, Bond et al.
[1992] reported 5- to 10,000-year intervals between periods of
marked decreases in sea surface temperature and salinity
(>10
C and 
1%
, respectively) which
are inconsistent with Milankovitch periodicities (23, 41, and
100 thousand year cycles) as a causal mechanism.
Sediments deposited in the lakes of Africa, central Europe,
and the southeastern United States also record slow oscillations
of the climate system. Sediments from the Sahel (Niger) and
Sahara (Algeria) deposited during the last deglaciation mark a
two-step transition from arid to humid conditions that was
synchronous in the two regions [ Gasse et al., 1990]. The
transition was inferred to be consistent with global changes in
ocean and atmospheric dynamics at that time because the phase of
maximum aridity in Africa falls within the cold Younger Dryas
chronozone of Europe. The transitions, recorded as 
O, are
approximately -7%
over 1-2 kyr. This corresponds
to roughly 5-10
C/kyr, assuming that 5%
O values are associated
with temperature changes of 
7
C as given by
Dansgaard et al. [1989]. Talbot and Johannessen [1992]
also reported the occurrence of major dry intervals in Africa
immediately following the Last Glacial Maximum (
18 kyr BP).
Their record revealed oscillations between shallow, brackish
water (i.e., dry conditions) lasting 
6 kyr during the
Pleistocene and deep, dilute, periodically overflowing conditions
lasting 4-10 kyr before and after the dry episode (
25 and 5
kyr BP). Rapid changes were observed in a study of loess input
to Lake Constance, in the Alps of central Europe [ Niessen et
al., 1992]. These lake cores showed a regular pattern of spring
deposition reflecting stable, arid, windy conditions of the Older
Dryas time period, followed by a 1,000-year transition to a
milder climate that preceded the onset of the Bø lling/Allerø d
climate amelioration by 
1 kyr. Pollen from Lake Tulane,
Florida, indicate shifts between pine forests ( Pinus)
during wetter conditions and open oak savanna ( Quercus)
during drier periods of the past [ Grimm et al., 1993].
Time series analysis of these data depicted a dominant
periodicity between the two primary vegetation types of
5,700 years and a less persistent 1,224-year frequency.
While different types of records (e.g., sea sediments versus ice cores) tend to highlight variability in the earth climate system over different timescales, the combination of available evidence has shown that the system varies appreciably on a variety of timescales, and that it is the overlapping of these cycles of variability that makes identification of patterns, and thus predictions of future climate, challenging.