|
||||||||||||||||
|
||||||||||||||||
The following HTML text is provided to enhance online readability. Many aspects of typography translate only awkwardly to HTML. Please use the page image as the authoritative form to ensure accuracy. Page 25 3
|
||||||||||||||||
 |
 |
Page 25 |
 |
 |
 |
|
| ||||||||||||||||||||||||||||||||
|
|
|||||||||||||||||||||||||||||||
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 25
Page 25
3
Modes of Climate Variability
Climate can be loosely defined as the ensemble of weather. As
such, it is inherent to the atmosphere, but is affected by
interactions between the atmosphere and the ocean, the biosphere,
the land surface, and the cryosphere. Because climate varies on all
time scales (see, e.g., Mitchell, 1976; NRC, 1995), an appropriate
mean can serve as a reference state for the study of variability on
shorter time scales, while itself changing on longer time scales.
In practice, whatever the definition used for the mean, an anomaly
is the difference between the instantaneous state of the climate
system and that mean. Climate variability and change are
characterized in terms of these anomalies. As the study of climate
progresses, it is becoming increasingly apparent that the
variations are not randomly distributed in space and time, but
often appear to be organized into relatively coherent spatial
structures that tend to preserve their shape—or assume a
limited number of related shapes—while their amplitude, phase,
and sometimes geographic position change through time. Though the
precise nature and shape of these structures, or patterns, vary to
some extent according to the statistical methodology employed in
the analysis, consistent regional characteristics that identify the
patterns still emerge. Therefore, in studying climate variability
and change, the examination of spatio-temporal patterns is a
natural development. Such study is also consistent with the IPCC
Second Assessment (IPCC, 1996a), where it was noted that much of
our attention in recent years has shifted from the analysis of
changes in mean global temperature to that of changes in the
spatial distribution of temperature and other climate variables,
reflecting the anticipation that climate may vary in both space and
time.
We do not yet have an exhaustive inventory of global and
regional patterns, nor do we understand their mechanisms,
relationships, temporal characteristics—such as persistence or
periodicity—or full implications for climate prediction.
Still, study of the most thoroughly investigated spatio-temporal
pattern—that which dominates the tropical Pacific and is
associated with the El Niño/Southern Oscillation (ENSO)
phenomenon—led to the first successful numerical climate
predictions, while yielding considerable insight regarding the
climate system, the nature of its air-sea couplings (Cane et al.,
1986) and its scales of teleconnectivity (NRC, 1996). Many of the
other patterns, while less well documented or studied, appear to
affect regional climate, as well as agricultural yields and
regional fish inventories, and appear to be related to the
frequency of hurricanes, variations in the ocean's thermohaline
circulation, and other things. These patterns vary over a broad
range of space and time scales, and their relative phasing can
dominate global and regional temperature variations. They often
show regional and global teleconnections, involve a number of
distinct climatological variables, and apparently focus different
forcings and processes into single coherent responses. Because of
these attributes and co-varying relationships, it is hoped that
their further study may ultimately yield benefits for dec-cen
climate predictability similar to those obtained for
seasonal-to-interannual predictions through the study of ENSO.
Spatio-temporal patterns thus provide one obvious avenue by which
the search for a predictable climate signal—that is, the
extraction from the complex climate system of a finite set of
regular components—should be pursued.
The literature is replete with descriptions of patterns,
covering a broad range of climatological variables and spatial and
temporal scales. Several of these patterns have received
considerable attention in recent years, and their names are now
firmly established in the climatological lexicon. The purpose of
this chapter is to provide a brief description of the more widely
discussed patterns that have been observed to vary on decadal or
longer time scales. This chapter thus serves as a glossary, albeit
incomplete, for the remainder of the text, while describing a
representative selection of patterns with their characteristics,
couplings, and relationships. A number of issues related to
improving our understanding of the role of spatio-temporal patterns
in climate change and variability over dec-cen time scales are
presented as well.
OCR for page 26
Page 26
Climate Patterns in the
Atmosphere
North Atlantic Oscillation
The North Atlantic Oscillation (NAO) is usually defined through
the regional sea-level pressure (SLP) field, although it is readily
apparent in mid-tropospheric height fields. Its influence extends
across much of the North Atlantic and well into Europe (Figure
3-1a). Like other patterns to be discussed here, it has a basically
fixed spatial structure. The NAO' s amplitude and phase vary over a
range of time scales from intraseasonal (van Loon and Rogers, 1978)
to interdecadal (Wallace et al., 1992); the largest amplitudes
typically occur in winter. Figure 3-1b shows more than 100 years of
NAO variability.
The NAO is often indexed by the difference in SLP between
Iceland, representing the strength of the Icelandic (or
Newfoundland) climatological low, and the Azores or Lisbon, near
the central ridge of the Azores high. Correlation of the NAO index
with surface air temperature and sea surface temperature (SST)
further reveals the extent of the atmospheric connection between
the North Atlantic and the northern portion of Europe, and part of
northern Asia (Hurrell and van Loon, 1996; Hurrell, 1995).
Typically,
Figure 3-1
(a) Differences between sea-level pressures in high and low NAO-index years, showing the region of
NAO influence. (b) Variation in the NAO (December-March) index since 1864; the heavy line represents
a filtered version of the data. (Both figures from Hurrell, 1995; reprinted with permission of the American
Association for the Advancement of Science.)
OCR for page 27
Page 27
when the index is high the Icelandic low is strong, which
increases the influence of cold Arctic air masses on the
northeastern seaboard of North America and enhances the westerlies
carrying warmer, moister air masses into western Europe in winter
(Hurrell, 1995). Thus, NAO anomalies are related to downstream
wintertime temperature and precipitation across Europe, Russia, and
Siberia (Hurrell and van Loon, 1996; Hurrell, 1995). They have also
been linked (see, e.g., Dickson et al., 1996) to changes in the
thermohaline circulation in the North Atlantic (Lazier, 1988), the
cod stock in the northwest Atlantic Ocean (Dickson et al., 1988),
the mass balance of European glaciers (Pohjola and Rogers, 1997),
the Indian monsoon (Dugam et al., 1997), and the atmospheric export
of North African dust (Moulin et al., 1997).
Pacific-North American Pattern
The Pacific-North American (PNA) pattern represents a
large-scale atmospheric teleconnection between the North Pacific
Ocean and North America. It appears as four distinct cells in the
500 hPa (hPa are equivalent to mb) geopotential height field near
Hawaii, over the North Pacific, over Alberta in Canada, and over
the Gulf Coast of the United States. Wallace and Gutzler (1981)
defined an index for the phase of this teleconnection pattern
through a weighted average of 500 hPa normalized-height anomaly
differences between the centers of the four cells. (Figure 3-2a
shows the region and extremes of influence of the PNA pattern, and
Figure 3-2b shows 15 years of variation in the PNA index.) The PNA
is reflected in SLP (Rogers, 1990) as well, however, and can
Figure 3-2
(a) Composite of the difference in the 500 hPa height field associated with the ten extreme positive and ten extreme
negative values of the PNA index for 1962-1977. (b) Time series of the monthly mean value of the PNA index. Only
Dec.-Jan.-Feb. values are shown, with the year tick marks at the Jan. values. (Both figures from Wallace and Gutzler,
1981; reprinted with permission of the American Meteorological Society.)
OCR for page 28
Page 28
be depicted by the North Pacific Index (NPI) (Trenberth, 1990).
The NPI is defined as the averaged SLP over a large area of the
North Pacific Ocean near the center of the Aleutian low.
It has been speculated that the decadal variability of the PNA
and NPI is associated with decadal changes in the tropical Pacific,
as discussed below (Nitta and Yamada, 1989; Graham, 1994; Lau and
Nath, 1996). Interdecadal variability of ENSO (discussed below in
the section on the role of dec-cen variability in global warming)
and the PNA is also thought to be responsible for a significant
amount of the variance in the salmon inventory along the northwest
coast of North America (Mantua et al., 1997).
Other Patterns of Interest
Regional Patterns
The NAO and PNA patterns introduced above, while predominant in
the literature and displaying variability on decadal time scales,
represent but two of those identified. A number of other regional
atmospheric or SST patterns have been analyzed, such as the North
Pacific Oscillation (NPO) (Walker and Bliss, 1932; Rogers, 1981),
West Pacific Oscillation, West Atlantic Pattern (Wallace and
Gutzler, 1981), and Pacific Decadal Oscillation (Mantua et al.,
1997). Other atmospheric patterns, such as the Pacific South
American pattern, have been identified in the Southern Hemisphere;
the data are frequently too sparse in time and space to allow more
detailed analyses of these patterns, however. In addition, other
structures exist that may or may not be considered climate
patterns, although they are often related to the other patterns or
presented in a similar manner. For example, the Asian monsoon,
while predominantly a seasonal signal, is strongly correlated with
ENSO and shows decadal variability as indexed by precipitation and
wind speeds over India. Some investigators treat it as another
distinct form of decadally varying pattern.
The number of regional patterns identified in the Northern
Hemisphere is on the order of 10 (Wallace and Gutzler, 1981;
Esbensen, 1984; Barnston and Livezey, 1987), which raises a
question as to whether they are all unique. If atmospheric
variability amounted to a continuum (i.e., no phase preference
existed, a situation comparable to "white noise" in the frequency
domain), then statistical analyses, such as those used in
teleconnection studies, would produce a finite number of
teleconnections (see, e.g., Wallace, 1996). Thus it has been
suggested that the multiplicity of patterns is partially the result
of a regional continuum (Kushnir and Wallace, 1989; Wallace, 1996),
and that not all teleconnections are indeed unique phenomena.
Cold Ocean-Warm Land
Finally, one other "pattern" warrants introduction here. This is
the "cold ocean-warm land" or COWL pattern (Wallace et al., 1995).
The COWL pattern is not a fundamental mode of climate
variability as defined through the decomposition of climatological
variable fields, nor is it a particular climate phenomenon; rather,
it simply represents a distinct geographic distribution of
near-surface temperature anomalies predominantly reflecting the
contrast in thermal inertia between land and ocean (Wallace et al.,
1995, Broccoli et al., in press). In particular, the COWL pattern
is a Northern Hemisphere winter phenomenon that is a manifestation
of the dominant effect of the continental land masses on the mean
surface air temperature during the cold season. It shows
considerable high-frequency variability (e.g., monthly), because
the air over the ocean responds slowly to change because the
ocean's large heat capacity makes it slow to change, whereas the
response of air over land is considerably faster because the land's
small heat capacity permits more rapid response. These differences
lead to rapid change over land in concert with large-scale shifts
in the atmospheric circulation cells, while change over the oceans
is much slower and more attenuated. Despite the apparent short-term
memory of the COWL pattern, it displays long-term variability as
well, which is of particular importance to the global warming
experienced over the last 20 years. The lower-frequency variability
of the COWL pattern seems to be related to long periods of
simultaneous surface warmth in northwestern North America and the
Eurasian continent, and can be identified with the simultaneous
phase-locking of the PNA and NAO patterns (Wallace et al., 1995;
Hurrell, 1996).
Co-Variability in the Climate System:
Coupled Patterns
Coupled patterns have expressions in at least two climate-system
components, but they are not presumed to be causally related. The
term therefore includes coupled modes, but also refers to patterns
in each component that are simply coherent.
Tropical Atlantic Variability
Numerous studies have found a robust relationship between SST
anomalies in the tropical Atlantic and changes in soil moisture,
albedo, and surface-roughness over North Africa (see Nicholson,
1989, for a brief overview). SST anomalies appear to be responsible
for a large part of the variability in tropical rainfall over
Africa and Brazil. Observational studies by Hastenrath and Heller
(1977) and Lamb (1978) were the first studies that linked the
variability in tropical Atlantic SST to variability in the rainfall
over the Sahel and Nordeste Brazil, respectively (see also Markham
and McLain, 1977; Lamb et al., 1986; Hastenrath and Greischar,
1993; Rao et al., 1993). (Figure 3-3 illustrates the correlation
between tropical Atlantic convection and Nordeste Bra-
OCR for page 29
Page 29
Figure 3-3
March-April field of convective activity in the tropical Atlantic sector.
Shown are the differences in the number of days per month with highly
reflective clouds for Nordeste Brazil wet (1984-1986, 1989) minus dry
(1980, 1982, 1983, 1990) years. Significance of differences at the 5 percent
level, as determined from t-test, is indicated by dot raster. (From Hastenrath and
Greischar, 1993; reprinted with permission of the American Geophysical Union.)
zil precipitation.) Together these observational studies
indicate that rainfall variability is associated with changes in
the position and structure of the Intertropical Convergence Zone
(ITCZ), a band of surface wind convergence characterized by strong
and frequent convective activity, which in turn is extremely
sensitive to variations in the meridional SST gradient. (See also
Lamb, 1978; Shinoda, 1990; Rowell et al., 1992; and Nobre and
Shukla, 1996, for discussion of this topic.) Similarly, atmospheric
general-circulation model (GCM) studies using prescribed SST
anomalies also show the importance of tropical SST anomalies in
generating and controlling rainfall anomalies in Brazil and Africa
(Moura and Shukla, 1981; Mechoso et al., 1990; Hastenrath and
Druyan, 1993; Folland et al., 1986; Owen and Folland, 1988; Palmer
et al., 1992; Rowell et al., 1992; Semazzi et al., 1993).
The tropical Atlantic Ocean shows a coherent structure in SST
variability. The dominant pattern of SST, as defined by empirical
orthogonal function (EOF) analysis, often shows a warm pool in the
tropical North Atlantic and a complementary cool pool in the
tropical South Atlantic, or vice versa. These centers of action
seem to vary coherently over decadal time scales but independently
on shorter time scales (Houghton and Tourre, 1992; Mehta and
Delworth, 1995; Chang et al., 1997). Consequently, this phenomenon
is sometimes referred to as the Atlantic Tropical Dipole, although
the lack of a clear consensus on the actual dipole nature of the
pattern leaves many simply referring to it as the decadal tropical
Atlantic SST variability. This low-frequency SST phenomenon shows
concurrent anomalies in the rainfall over Brazil and northern
Africa (Figure 3-4a). Periods of greater-than-normal rainfall were
experienced over northeast Brazil in the 1960s, and periods of
lower-than-normal rainfall in the mid-1970s to early 1980s (Figure
3-4b). Gray (1990) suggests that the decadal changes in the SST in
the subtropical North Atlantic may also be responsible for the
changes in the distribution and intensity of hurricanes in this
region.
While the physics of the dipole climate oscillations in the
Figure 3-4
(a) The spatial pattern of the EOF of Atlantic SST that is strongly related to rainfall in Nordeste Brazil and western Africa. (b) Time series
(solid line) of the March-to-May values of the EOF shown in (a) and the north Nordeste Brazil rainfall anomalies (dashed line). (From Ward
and Folland, 1991; reprinted with permission of John Wiley and Sons, Ltd.)
OCR for page 30
OCR for page 31
OCR for page 32
OCR for page 33
OCR for page 34
OCR for page 35
OCR for page 36
OCR for page 37
OCR for page 38
Representative terms from entire chapter:
cowl pattern
Page 30
tropical Atlantic are not yet understood, recent results from
Chang et al. (1997) demonstrate that an intermediate-level coupled
atmosphere-ocean model of the tropical Atlantic does support
decadal oscillations that have a structure similar to the observed
dipole phenomenon. Finally, the climate in and around the Atlantic
is also affected by variability in the tropical Pacific on both the
interannual (see, e.g., Folland et al, 1986; Hastenrath et al,
1987; Enfield and Mayer, 1997) and decadal (Zhang et al., 1997)
time scales.
As in the tropical Pacific, within 5º of the equator (the
so-called tropical waveguide), ocean dynamics seem to play a
significant role in the generation of SST anomalies on the
interannual time scale and in the equatorial waveguide (Zebiak,
1993). The SST anomalies in the waveguide are linked to interannual
variations in rainfall along the Guinea coast (Wagner and da Silva,
1994). However, little is known about the cause of the
low-frequency variability in the SST of the tropical and
subtropical Atlantic that is associated with the large-scale
rainfall anomalies and with the variations in hurricane
activity.
North Atlantic Variability
Kushnir (1994) examined the multidecadal variability in the
observational record of SLP, SST, and surface wind velocity in the
North Atlantic basin, and found two examples of warm and cold
epochs in the twentieth century; each of these epochs lasted more
than a decade. Warm periods are characterized by positive SST
anomalies around southern Greenland and negative anomalies along
the northeastern U.S. seaboard (upper panel of Figure 3-5). The
concurrent SLP and wind anomalies indicate a southern displacement
of the Icelandic low and a relaxation of the winds in the
subtropics (lower panel of Figure 3-5) coinciding with a decrease
in NAO. Kushnir (1994) concluded that the variability demonstrated
on these time scales in the observational record was governed by a
basin-scale interaction between the large-scale oceanic circulation
and the atmosphere.
Deser and Blackmon (1993) found that SST in the North Atlantic
subpolar basin varied concurrently with the atmospheric surface
wind anomalies. These atmosphere and ocean anomalies span the
twentieth-century record and display a roughly 10-year period.
Deser and Blackmon also note that the spatial relationship between
SST and wind anomalies in these quasi-decadal cycles is consistent
with what could be expected theoretically if the phenomenon were
inherently due to coupling between the atmosphere and the ocean.
They point out, however, that there is a high negative correlation
between the SST anomalies and the anomalies in sea-ice extent in
the Baffin Bay/Labrador Sea region. Furthermore, the sea-ice
anomalies lead the SST and wind anomalies by a few years.
In addition to the above phenomena, an event that began in the
late 1960s has drawn unusual attention from the ocean community. A
significant surface freshwater anomaly ap-
Figure 3-5
Upper panel: The difference between the annual winter Atlantic SST
averaged from 1950-1964 (warm years) minus the winter average from
1970-1984 (cold years). Contour interval is 0.2ºC. Lower panel: as above,
but for SLP and winds. Contour interval is 0.5 mb. The arrow at the
bottom of the panel is 1 m s-1. Distribution of the t-variable corresponding
to SST and SLP differences is denoted in three levels of gray: light for 2.0-
2.5, medium for 2.5-3.0, and dark for 3.0-3.5. (From Kushnir, 1994; reprinted
with permission of the American Meteorological Society.)
peared in about 1969 in the Labrador Sea. Now known as the
"Great Salinity Anomaly" (GSA), this feature can be traced moving
eastward across the subpolar gyre, into the Norwegian Sea, and
ending up near Fram Strait more than 10 years later (Dickson et
al., 1988). Aagaard and Carmack (1989) hypothesize that the GSA was
born from an increase
Page 31
Figure 3-6
Global SST pattern, wind stress, and sea-level pressure that is related to the interannual variability associated with ENSO,
based on linear regression between a high-pass-filtered "cold-tongue index" (CT) and global SST. (From Zhang et al.,
1997; reprinted with permission of the American Meteorological Society.)
Figure 3-7
Time series of a cold-tongue index, corresponding to the SST pattern displayed in Figure 3-6.
(From Zhang et al., 1997; reprinted with permission of the American Meteorological Society.)
Page 32
Figure 3-8
Global residual (GR) SST pattern, wind stress, and sea-level pressure, from which the linearly related ENSO
variability has been removed. (From Zhang et al., 1997; reprinted with permission of the American Meteorological Society.)
Figure 3-9
Time series of the global-residual index, corresponding to the SST pattern displayed in Figure 3-8,
which indicates that the ENSO-like pattern in that figure is associated primarily with decade-to-
century-scale variability. (From Zhang et al., 1997; reprinted with permission of the American
Meteorological Society.)
Page 33
in sea-ice export from the Arctic (see also Hakkinen, 1993, and
a related paper by Mysak et al., 1990). There is evidence that the
GSA's stabilization of the upper water column interrupted
deep-water production in the North Atlantic (Lazier, 1988). Thus,
the GSA is thought to be an example of phenomena that may lead to
changes in the thermohaline circulation in the ocean, which may
then lead to feedbacks to the atmosphere on much longer time
scales.
Pacific Decadal Enso-like Pattern
Tanimoto et al. (1993) and Zhang et al. (1997) demonstrate that
the time variability of the leading EOF of the global SST field is
separated into two components: one identified with ENSO variability
(Figure 3-6) showing predominantly interannual variability (Figure
3-7), and the other a linearly independent "residual" (Figure 3-8)
dominated by dec-cen-scale variability (Figure 3-9). The two
components exhibit remarkably similar spatial signatures in global
SST, SLP, and wind-stress fields, with the SST field in the
residual pattern being less equatorially confined in the eastern
Pacific than the interannual pattern, and having a larger
extratropical signature in the North Pacific. In fact, the residual
pattern is very similar to the leading EOF of North Pacific SST
that Mantua et al. (1997) have called the Pacific (inter)Decadal
Oscillation (PDO). The residual pattern's SLP signature is also
stronger in the extratropical North Pacific, and its wintertime 500
hPa height anomaly (Figure 3-10) more closely resembles the PNA
pattern than the ENSO pattern shown in Figure 3-6. The amplitude
time series of the ENSO-like pattern in the residual variability
reflects much of the low-frequency variance in the data as well as
some of the interannual variability, including a notable regime
shift in 1976-1977 (Quinn and Neal, 1984) and an equally remarkable
shift of polarity in the 1940s (see Figure 3-9).
This ENSO-like pattern in SST appears to be teleconnected to
anomalies in the mid-latitude atmosphere and ocean of the North
Pacific (Figure 3-6). The decadal ENSO-like anomalies of Figure 3-8
are also teleconnected throughout the tropics, with large
concurrent changes in tropical Atlantic and Indian Ocean SST (Zhang
et al., 1997), as well as the North Pacific (see, e.g., Kumar et
al., 1994). Unlike the previous patterns, which are clearly defined
through EOF analyses of atmospheric or ocean property fields, this
pattern appears only when the data (atmosphere or ocean fields) are
time-filtered prior to the analysis. However, because of its large
spatial influence and its apparent relationship to the
shorter-time-scale ENSO phenomenon, it has received considerable
attention in recent years.
The last few decades have represented a warm phase of this
climate anomaly, which has preceded a significant reduction in the
alpine glaciers throughout the tropics (Thompson et al., 1993b;
Diaz and Graham, 1996). In addition, the streamflow and snowpack in
the northwest and southwest of North America (Cayan and Peterson,
1989; Cayan, 1996) are well correlated with this time series of the
decadal ENSO-like climate phenomenon.
Figure 3-10
The 500 hPa height anomalies associated with the cold-tongue index (CT) (right panel), characterized by interannual
ENSO variability, and the global-residual index (GR) (left panel), characterized by ENSO-like decadal variability.
(From Zhang et al., 1997; reprinted with permission of the American Meteorological Society.)
Page 34
Pacific Summer Variability
Norris and Leovy (1994) point out that there are long-term
trends in summer SST in the North Pacific over the period
1952-1981, and that the changes in SST are significantly
anti-correlated with the long-term trends in the maritime
stratiform cloudiness. They note that the strongest
(anti)correlations between SST and maritime stratiform cloudiness
are found co-located with the largest climatological SST gradients.
Norris and Leovy suggest that these trends may result in part from
the persistence of SST anomalies from winter to summer.
Arctic Variability
When Walsh et al. (1996) examined the long-term record of SLP in
the Arctic basin, they found that a large decrease in the annual
mean SLP occurred in the mid-to-late 1980s. The anomalously low SLP
persisted through 1994, the end of the record they examined. The
SLP anomalies are largest in the central Arctic, decreasing toward
the adjacent coastal regions around the Arctic Ocean; they are
present year-round but are greatest in the winter season. The
annual mean pressure changes are larger there than anywhere else in
the North-em Hemisphere.
Walsh et al. argue that it is unlikely that the observed change
in the Arctic atmospheric circulation could be associated with
low-frequency variability in the extrapolar regions. They note,
however, that the observed changes in the atmospheric circulation
can be expected to lead to changes in the transport and compactness
of the sea ice in the Arctic basin. Specifically, the mean
anticyclonic motion of the ice pack should decrease, and the sea
ice should now be more divergent than it was during the 1970s and
early 1980s. Indeed, there have been two remarkable large-scale
anomalies in Arctic sea ice in the past decade: the extraordinarily
thin sea ice that was experienced by the SHEBA (Surface Heat Budget
of the Arctic project; see Moritz and Perovich, 1996, for a
description) in the winter of 1997 (McPhee et al., 1998), and the
offshore contraction of the sea ice off Siberia in 1990. The
latter, which is unprecedented in the record, has recently been
linked to changes in the pan-arctic atmospheric circulation
(specifically, the wind stress) by Serreze et al. (1995) and Bitz
(1997). Cavilieri et al. (1997) note that the areal extent of sea
ice has decreased by about 6 percent between 1978 and 1996.
The large trends in the Arctic atmospheric circulation and in
the NAO (see, e.g., Figure 3-1) during the past 30 years appear to
be related (see Thompson and Wallace, 1998, and references
therein). Thompson and Wallace argue that these large-scale decadal
trends are best described in terms of a planetary-scale mode of
variability, which they referred to as the Arctic Oscillation (AO),
whose regional extension into the North Atlantic accounts for the
phenomena attributed to the NAO. The AO involves fluctuations in
the strength of the polar vortex that extend from the surface
upward into the lower stratosphere, and occur on time scales
ranging from weeks to decades. The ''high index'' (strong
westerlies) polarity of the AO is characterized by mild wintertime
temperatures over most of Eurasia poleward of 40ºN.
Stratospheric involvement in the AO is most clearly apparent during
late winter and early spring, when wave-mean-flow interactions at
these levels are strongest. Thus, the trend in the AO, and hence in
the NAO and Arctic circulation, may be viewed as a systematic bias
in one of the atmosphere's most prominent modes of internal dynamic
variability. Whether it is occurring in response to anthropogenic
forcing has yet to be determined.
Thompson and Wallace (1998) found that the AO, as represented by
the first empirical orthogonal function (EOF) of SLP, changed
rapidly in amplitude during the mid- to late 1980s. This change is
consistent in timing and sense with the changes noted by Walsh et
al. (1996). Furthermore, this atmospheric change coincides with a
number of additional changes noted in the sea ice and upper ocean.
(The precise timing of the changes is unknown for most of the
upper-ocean variables, given the relatively sparse data available
for the Arctic region.) For example, data collected through the
early 1990s by Morison et al. (1998), Carmack et al. (1995, in
press), McLaughlin et al. (1996), and Steele and Boyd (in press)
all show that the position of the central Arctic upper-ocean front,
which separates the Atlantic and Pacific waters, has shifted
relative to the 1950-1989 climatological position (as established
using the Environmental Working Group atlas built from U.S. and
Russian hydrographic observations released through the
Gore-Chernomyrdin joint commission); the region dominated by
Atlantic water has expanded by nearly 20 percent. These same data,
as well as those of Anderson et al. (1994), Rudels et al. (1994),
and Quadfasel (1991) suggest a warming of the Atlantic layer
occurring at this same time. Other studies are finding changes in
the surface winds, sea ice, and other upper-ocean characteristics,
such as pycnocline properties. Four summers with the most extreme
minimum Arctic sea-ice coverage (Maslanik et al., 1996) have
occurred since this change took place, and anomalously thin ice has
also been reported (McPhee et al., 1998). Together, these ocean,
ice, and atmosphere observations suggest that the changes in the
late 1980s in the Arctic may have involved the entire vertical
column from the upper ocean to the stratosphere.
There are, however, theoretical reasons to expect large
variability in Arctic systems because of the coupling among the
polar atmosphere, ocean, and sea ice. First, the Arctic sea-ice
thickness is thought to be extremely sensitive to changes in
vertical heat transport in the ocean (see, e.g., Maykut and
Untersteiner, 1971); only modest circulation changes in the ocean
would be required to induce variability in the ocean heat transport
that would have a significant impact on the thickness and spatial
extent of the Arctic sea ice,
Page 35
and hence on the albedo of the Arctic. In addition, it has
recently been argued that the observed high-frequency (subseasonal)
variability in atmospheric energy transport into the Arctic may
lead to large variability in Arctic sea-ice thickness in the Arctic
that occurs primarily on decadal and multidecadal time scales (Bitz
et al., 1996).
Variability of the sea ice in the Arctic is a defining aspect of
the Arctic climate system. In addition, the ramifications of
changes in Arctic circulation and sea ice are clearly important for
understanding the circulation of the subpolar North Atlantic Ocean
mentioned earlier. Thus, while the connections between the AO and
NAO, and the polar and extra-polar regions, are still unclear,
their co-variability suggests that these coupled Arctic variations
are intimately tied to extra-polar regions.
Antarctic Variability
A completely different kind of pattern involving sea ice,
surface winds, SST, and SLP has been found in the Southern Ocean.
Specifically, the Antarctic Circumpolar Wave (ACW) is characterized
by co-varying deviations in monthly climatological averages of
these variables along the Antarctic polar front, near the winter
marginal ice zone (White and Peterson, 1996). It is also highly
coherent with temporal variations in ENSO (White and Peterson,
1996) and the Indian Ocean monsoons (Yuan et al., 1996), although
the underlying physics are not yet understood. It is predominantly
an interannual phenomenon, but, as with ENSO, it shows
longer-period variability. It is not clear how the ACW is related
to, or interacts with, the dominant mode of variability of the
zonal mean flow in the Southern Hemisphere (Hartmann and Lo, 1998),
or the standing-wave patterns of van Loon and Jenne (1972) and van
Loon et al. (1973), though the superposition of these various
patterns in space suggests that their interaction is feasible.
The Role of Dec-Cen Variability in
Global Warming
It is clear that the global warming experienced over the past 20
years is distinguished by an enhanced warming in winter that was
not evident in previous decades, dominated by a strong warming over
Northern Hemisphere land, and compensated for to some degree by a
lesser cooling over parts of the Northern Hemisphere oceans.
Despite its decadal persistence, this pattern is consistent with
the basic COWL pattern described above. Its geographic aspect is
readily apparent in the global surface-temperature data when the
last 20 years are compared to the previous 20 years (Color plate
1). A pattern similar to the COWL pattern (though not identified as
the COWL pattern) is produced by numerous modeling studies
simulating anthropogenic inputs, and thus is considered by some to
be one component (of several) of the so-called "greenhouse
fingerprints" (Wigley and Barnett, 1990; Santer et al., 1995). Its
presence in the actual observational data has therefore been
accepted as additional evidence of anthropogenic warming (IPCC,
1996a).
Figure 3-11 shows that when the monthly-average Northern
Hemisphere surface-temperature time series for this century is
adjusted to eliminate the influence of the COWL pattern, two things
become apparent (Wallace et al., 1995): (1) a large fraction of the
month-to-month variability, which is particularly apparent in the
colder half of the year, is removed; and (2) a significant fraction
of the accelerated warming in the cold-season months that has been
apparent since the mid-1970s (Figure 3-11, middle panel) is also
removed, and the summer and winter trends become comparable (bottom
set of curves).
Further investigation suggests that much, though not all, of the
accelerated warming that has occurred since the mid-1970s is
attributable to the COWL pattern. Much of the COWL pattern itself
can be explained by the synchronous polarity of the NAO and PNA
patterns during this accelerated period of warming (Hurrell, 1996).
That is, over the
Figure 3-11
Top, smoothed monthly-average surface-temperature time series for the Northern Hemisphere.
The solid line is the warming-months average, the dashed line the cooling-months average.
Center, the portion of the series attributed to the COWL pattern. Bottom, the time series adjusted
by eliminating the COWL contribution. (From Wallace et al., 1995; reprinted with permission of the
American
Association for the Advancement of Science.)
Page 36
last 20 years, the NAO and PNA patterns (the latter represented
by the North Pacific Index) both seem to show a seemingly unusual
persistent tendency, on average, to occupy states that favor a
warming of Europe and Northern Asia by the NAO, and warming of
North America by the PNA. When this warming is removed, the global
warming trend of the last 25 years is similar to, though slightly
larger than, the warming that occurred over several decades in the
early part of this century (from 1910 to 1940, say). Broccoli et
al. (in press) find that their coupled model reproduces this
warming attributed to the COWL pattern. They also find that such an
accelerated warming is highly anomalous (exceeding the 99th
percentile) in simulations that do not contain greenhouse warming.
The extent to which the COWL pattern is anthropogenically caused or
influenced is an outstanding issue, and one crucial to all aspects
of the greenhouse-warming puzzle.
Another critical question is whether anthropogenic increases in
radiative forcing play a role in amplifying the intensity,
frequency, and/or duration of ENSO events. For example, several
works have shown that the coupled ocean-atmosphere system,
particularly in the tropical Pacific Ocean, plays an important role
in regulating the mean climatological state (see, e.g., Neelin and
Dijkstra, 1995; Sun and Liu, 1996; Clement et al., 1996; Seager and
Murtugudde, 1997; and Cane et al., 1997). Sun and Trenberth (1998)
extend this concept by suggesting, on the basis of an observational
study, that the E1 Niño events are an effective means for
removing heat from the tropical Pacific and may even arise as a
necessity for removing excess heat associated with increased solar
heating and the greenhouse effect. This suggestion is also
consistent with a number of coupled-modeling studies of increased
radiative forcing or increased CO2
that have shown more energetic and more extreme ENSO events (Sun,
1997; Timmermann et al., 1998; and an ENSO-like pattern of response
(see, e.g., Meehl and Washington, 1996, and Knutson and Manabe,
1995, in press). Much of our current uncertainty regarding the role
of the coupled ocean-atmosphere system in greenhouse warming is the
result of uncertainties involving the role of the cloud-radiative
feedback, which underscores the need to improve model
representations of this critical process. Substantial uncertainties
regarding model response of the tropical Pacific to greenhouse
warming also arise in the modeling of the equatorial
thermocline.
Such a response has been suggested by Trenberth and Hoar (1995,
1997). Through analyses of more than 100 years of climate data they
concluded that the record of the Southern Oscillation Index (SOI)
in the period since 1976 is significantly different from the
earlier portion of the record. Though this research preceded the
powerful 1997-1998 El Niño event, they found a recent
tendency for more frequent El Niño events and fewer La
Niña events (contrary to the model results of Timmermann et
al. (1998)). Although Trenberth and Hoar posit that this change may
be caused by the observed increase in greenhouse gases, it is not
currently possible to say definitively what underlies this change
in character of El Niño events (NRC, 1996). Trenberth and
Hoar's conclusions are predicated on a statistically stationary,
linear-time-series model of the SOI time series. Rajagopalan et al.
(1997) offer an alternative viewpoint using a time-series model
whose parameters are allowed to vary slowly over time. This
analysis highlights significant dec-cen variability both in the
probability of El Niño and La Niña states, and in the
probability of transitions between these states. Their results
suggest that the recent tendency to more frequent and persistent El
Niño events may not be nearly so unusual as Trenberth and
Hoar indicate. These differences underscore the need to
discriminate better between natural and anthropogenic causes for
dec-cen variability, and to think further about what statistical
methodology is appropriate for such analyses. By improving our
understanding of how anthropogenic climate change interacts with
natural processes such as El Niño, we will improve our
ability to detect global warming signals, and perhaps gain
predictive insight into the dec-cen-scale modulation of phenomena
that have traditionally been viewed in a seasonal-to-interannual
context.
Finally, it is interesting to note that the analysis of Huang et
al. (1998) of 135-year-long indices has revealed a relationship
between the NAO, ENSO, and PNA, albeit a complex one. They find
that an enhanced positive phase of the PNA is likely to occur with
the positive phase of the NAO. Furthermore, at frequencies of 2 to
4 years, the NAO is coherent with ENSO from 1960 to 1990; this
coherence is most apparent during strong to moderate El Niño
events. (This result is consistent with the findings of Rogers
(1984).) As noted earlier, the COWL spatial pattern also correlates
with positive NAO and PNA indices. Greenhouse warming thus carries
the potential for altering all four patterns. Our understanding of
the climate system' s likely response to anthropogenic forcing
might benefit from considerable attention to the relationships
between greenhouse warming and these (and other) natural modes of
variability.
Fundamental Issues and Questions
The patterns and coupled modes occupy large spatial areas,
describe significant climate variance, and bridge high-, mid-, and
low-latitude zones. Despite the uncertainties about their roles in
anthropogenic global warming and natural climate change, they
represent an obvious avenue through which coherent climate
variations and change may be propagated globally. The IPCC Second
Assessment (IPCC, 1996a) noted that much of our attention has
recently shifted from the analysis of mean global temperature to
that of its spatial distributions, anticipating that climate change
might manifest itself irregularly in space and time—yet
patterns have appeared. The identification of coherent patterns
with coupled modes that explain significant fractions of the
spatial and temporal variability offers hope that there may be
a
Page 37
signal in what appears to be just "noise." Furthermore, the
apparent persistence of these patterns in time, even allowing for a
slow evolution, provides additional hope that these signals may be
exploited to help us understand and predict future climate
variability and change. Also, the apparent relationship between
specific climate-pattern dispositions and regional climate
characteristics lends support to the notion that understanding
long-term trends in the patterns may enable us to make short-term
predictions for some regions.
To realize these potentials, considerable effort must also be
invested in improving our general understanding of the patterns and
coupled modes: their mechanisms (dynamic and thermodynamic, natural
and anthropogenic), couplings, feedbacks, and sensitivities. These
are truly cross-disciplinary issues, requiring a strong
interdisciplinary approach. Specifically, we must answer the
following questions:
• What is the longevity of the patterns and their
spatial/ temporal variance? The patterns offer tantalizing
evidence that some fraction of the Earth's climate shows spatially
and temporally coherent structure with some degree of (predictable)
persistence in a time-averaged sense. However, the fundamental
patterns themselves may be transitory phenomena reflecting the
current configuration of a slowly changing climate. Consequently,
we need to understand the dynamics between climate patterns and the
general state of climate. For example, do the patterns and climate
evolve in a systematic manner? At some level of change do the
patterns become simple artifacts of the change, as opposed to
predictors of the change?
• What is the best way of characterizing the known
patterns, and are there additional patterns of interest?
Specifically, how can we most effectively define their salient
features, their co-varying components and coupled modes (including
regional influences and correlation with or control of the climate
attributes discussed in Chapter 2), their sensitivities to analysis
technique, and their spatial distribution and broader
teleconnections? Likewise, robust and optimal indices of these
coherent atmospheric-circulation patterns need to be established,
because some of the indices employed to represent the patterns,
while convenient, do not capture much of their spatial and temporal
complexity. For example, the Bermudan-Azorean high remains
relatively stable in its spatial orientation, but the Icelandic low
often migrates southward to Newfoundland, and the North Atlantic
SST pattern tends to show a rotation around the North Atlantic
basin (see, e.g., Hansen and Bezdek, 1996) that a simple dipole
index, such as the NAO, relating two fixed points, cannot capture.
The new findings of Thompson and Wallace (1998), showing that the
first EOF of Arctic SLP may be an even better indicator of Eurasian
climate change than the NAO, further underscores the need to
examine the optimal means of classifying the relevant modes of
climate. Therefore, while indices have proven useful in their
ability to simplify the temporal history of complex patterns and
demonstrate their broad spatial coherence and importance,
additional research is required to better characterize the patterns
and isolate their significant characteristics. That is, more robust
indices that efficiently describe the fundamentally important
characteristics of the patterns must be developed. What patterns
and coupled modes exist in data-poor regions, and what are their
spatial and temporal characteristics?
• Which patterns represent true dynamic modes, and which
ones are simply statistically consistent structures, or
geographically forced distributions? That is, are the patterns
fundamental modes of climate variability reflecting coupled,
internal, and external dynamics and thermodynamics? Or are they
simply the reflection of structures that are intrinsic to the
atmosphere (i.e., determined by the land-sea distribution and
internal atmospheric dynamics)? In the latter case, they
stochastically force the ocean (see, e.g., Hasselmann, 1976),
integrate the response, and provide a feedback to the atmosphere by
"reddening" the spectrum of the patterns in the atmosphere without
affecting atmospheric dynamics (Barsugli and Battisti, 1998). Or
are they the consequence of statistics or chaos, representing
attractors of random but spatially consistent distributions?
Understanding the mechanisms underlying these patterns will be
fundamental in our assessment of how they can ultimately be used in
long-term forecasting and prediction of climate variability and
change.
• What are the mechanisms responsible for generating,
maintaining, and modifying the patterns? What role do these
mechanisms play in the spatial propagation of regionally initiated
variability or change, and what are their critical dependencies
? To understand how a change in the state of a pattern in one
location may dictate the regional climate in some more remote
location, it is necessary to understand the mechanisms that control
the spatial and temporal evolution of the patterns and their
broader influences or teleconnections. This knowledge will also
provide an indication of how a local disturbance may influence the
dominant regional pattern, leading to broader propagation of the
anomaly, and thus influencing the controlling components of the
climate system.
• What is the relationship between decadal-to-centennial
patterns and global warming? What part of the COWL warming
pattern is due to natural variability versus anthropogenic
modulation of these naturally occurring patterns? Will there be
extended periods of time in which they display similar, relatively
persistent polarity quite by chance, or is COWL a manifestation of
anthropogenic warming through the polarity of the natural climate
modes? (See discussions by Wallace et al. (1995) and Hurrell
(1996).) Is the residual warming—that is, warming apart from
the COWL contribution—natural variability or anthropogenic
warming? What is the relationship between the COWL pattern,
greenhouse fingerprint, and natural climate patterns—that is,
how do the natural modes of the climate system respond to different
changes in forcing, natural or anthropogenic? Are there unique
characteristics or
Page 38
response modes? What controls the degree and nature of the
spatial distributions and interactions of the modes, and how do
they co-vary?
It is clear that a more comprehensive understanding of the
variability of these patterns on decade-to-century time scales is
absolutely essential if we are truly to distinguish between
anthropogenic change and natural climate variability, or understand
their interaction. If an enhanced greenhouse effect strengthens or
phase locks existing natural modes of variability, the study of
dec-cen variability and the study of anthropogenic change are
inextricably combined. It will be impossible to understand one
without the other.