date: Thu, 12 Oct 2000 12:12:38 -0400
from: "Raymond S. Bradley" <rbradley@geo.umass.edu>
subject: synthesis section 1-revised
to: k.briffa@uea.ac.uk, jcole@geo.Arizona.EDU, mhughes@ltrr.arizona.edu

<x-flowed>
Can you take a quick glance through this.  I've revised it & reorganised 
things a bit--regional sections folow this
Ray


Chapter 6.  The Climate of the Last Millennium
Raymond S. Bradley, Keith R. Briffa, Julia E. Cole and Malcolm K. Hughes

6.1  Introduction
We are living in unusual times.  Twentieth century climate was dominated by 
near universal warming; almost all parts of the globe had temperatures at 
the end of the century that were higher than when it began, and in most 
areas temperatures were significantly higher (Parker et al., 1998; Jones et 
al., 1999; Wallace et al., 2000).  [Do we want a figure here  Figure 6.1say 
a global map of (1900-1909)-(1990-1999) MAT from instrumental 
data?]  However the instrumental data provide only a limited temporal 
perspective on present climate.  How unusual was the last century when 
placed in the longer-term context of climate in the centuries and millennia 
leading up to the 20th century?  Such a perspective encompasses the period 
before large-scale contamination of the global atmosphere and global-scale 
changes in land-surface conditions.  By studying the records of climate 
variability and forcing mechanisms in the recent past, it is possible to 
establish how the climate system varied under natural conditions, before 
anthropogenic forcing became significant.  Natural forcing mechanisms will 
continue to operate in the 21st century, and will play a role in future 
climate variations, so regardless of how anthropogenic effects develop it 
is essential to understand the underlying background record of forcing and 
climate system response.
Sources of information on the climate of the last millennium include: 
historical documentary records, tree rings (width, density); ice cores 
(isotopes, melt layers, net accumulation, glaciochemistry); corals 
(isotopes and other geochemistry, growth rate); varved lake sediments 
(varve thickness, sedimentology, geochemistry, diatom and pollen content); 
banded speleothems (isotopes).  These are all paleoclimatic proxies that 
can provide continuous records with annual resolution.  Other information 
may be obtained from sources that are not continuous in time, and that have 
less rigorous chronological control.  Such sources include geomorphological 
evidence (e.g. from former lake shorelines and glacier moraines) and 
sub-fossil macrofossils that indicate the range of plant or animal species 
in the recent past.  In addition, ground temperature measurements in 
boreholes reflect the past history of surface temperatures, with temporal 
resolution decreasing with depth.  These provide estimates of overall 
temperature changes from one century to the next (Pollack et al. 19XX).
Proxies of past climate are generally controlled by a particular aspect of 
the system that causes a climate-related signal to be recorded.  For some 
biological proxies, such as tree ring density or coral growth rate, the 
main factor might be temperature or more specifically, the temperature of a 
particular season (or even just part of a season).  Density and growth rate 
might also be influenced by antecedent climatic conditions, or by other 
non-climatic factors.  Similar issues are important in other proxies, such 
as the timing of snowfall events that make up an ice core, or the rate and 
timing of sediment transport to a lake.  Though we recognize that the 
details of such relationships are important, proxies are rarely interpreted 
directly in terms of very specific controls, but rather in terms of 
temperature or precipitation in a particular season.  In many cases the 
main climatic signal in a proxy record is not temperature alone.  For 
example, evidence of a formerly high lake level may indicate higher 
rainfall amounts and/or a decrease in evaporation related to cooler 
temperatures.  Such issues are grist to the paleoclimatologists mill and 
are the subject of numerous studies.  Suffice it to say that proxies are 
generally selected to optimize a reconstruction of either temperature or 
precipitation and it is these studies that provide the basis for our review.
Changes in temperature have large-scale spatial coherence, making it easier 
to identify major variations with relatively few records.  Precipitation 
changes are more local or regional in extent, but they often reflect 
circulation changes that may have large-scale significance (as, for 
example, in ENSO-related rainfall increases that commonly occur in the 
southeastern U.S. during strong El Nio events; Stahle et al., 1998).  In 
this chapter, we focus mainly on temperature variations, but precipitation 
and hydrological variability are examined where there is good evidence for 
important changes at the regional scale.  In particular, we ask two 
questions regarding each attribute:
       does the 20th century record indicate unique or unprecedented 
conditions?
       do 20th century instrumental data provide a reasonable estimate of 
the range of natural variability that could occur in the near future?
First we deal with the overall pattern of temperature change at the largest 
(hemispheric) scale.  Then, we examine variability over four large 
sub-regions of the globe.  These results are placed in the longer-term 
perspective of Holocene climatic changes, and finally we examine forcing 
factors that may have played a role in the variations that have been 
identified.

6.2  Temperatures over the last millennium
Most high resolution paleoclimate records (i.e. those with annual 
resolution and a strong climate signal) do not extend back in time more 
than a few centuries.  Consequently, while there are numerous paleoclimate 
reconstructions covering the period from the 17th century to the present, 
the number of high resolution millennium scale records is very 
limited.  Continuous records are restricted to ice cores and laminated lake 
sediments, where the climatic signal is often poorly calibrated, and to a 
few long tree ring records, generally from high latitudes.  Inevitably, 
this leads to large uncertainties in long-term climate reconstructions that 
attempt to provide a global or hemispheric-scale perspective.  Bearing this 
in mind, what do current reconstructions tell us about the last millennium?
Figure 6.2 shows a reconstruction of northern hemisphere mean annual 
temperature for the last 1000 years, based on a set of over [100?] 
well-distributed paleoclimatic records spanning the past 600 years but a 
much smaller number of data sets (12) for the period prior to A.D. 1400 
(Mann et al., 1998, 1999).  The paleoclimatic proxies were calibrated in 
terms of the main modes of temperature variations (eigenvectors) 
represented in the instrumental records for 1902-1980.  Variations across 
the network of proxies, for the period before instrumental records, were 
then used to reconstruct how the main temperature patterns (i.e. their 
principal components) varied over time.  By combining these patterns, 
regional, hemispheric or global mean temperature changes, as well as 
spatial patterns over time were reconstructed (Mann et al., 2000).  To 
accurately reproduce the spatial pattern requires that the proxy data 
network is extensive enough to capture many of the principal eigenvector 
patterns.  With the data available, regional patterns of temperature 
variation could only be meaningfully reconstructed for 250 years, although 
the large-scale (hemispheric) mean temperature could be reconstructed for a 
longer period.  This is possible because the proxy data network, even at 
its sparsest, exhibits a coherent response to variability at the largest 
scale.  Thus a reconstruction of hemispheric mean temperature back 1000 
years is possible, using a quite limited network of data, albeit with 
ever-increasing uncertainty (i.e. expanding confidence limits) the further 
back in time one goes (Figure 6.2Mann et al 1000yr reconstruction with 
errors).  This reconstruction shows an overall decline in temperature of 
~0.2Cfrom A.D. 1000 until the early 1900s when temperatures rose 
sharply.  Superimposed on this decline were periods of several decades in 
length when temperatures were warmer or colder than the overall 
trend.  Mild episodes, lasting a few decades, occurred around the late 
10th/early 11th century and in the late 13th century, but there was no 
period when mean temperature was comparable with levels in the late 20th 
century.  Coldest conditions occurred in the 15th century, the late 17th 
century and in the entire 19th century.
Other attempts to assess northern hemisphere temperatures have taken a 
simpler approach, either averaging together normalized paleo-data of 
various types (Bradley and Jones, 1993; Jones et al., 1998) or averaging 
data scaled to a similar range (Crowley and Lowery 2000).  Such approaches 
do not provide an estimation of uncertainty, and indeed may lead to rather 
arbitrary combinations of very diverse data (often having different 
temporal precision).  Nevertheless, the resulting time series from all of 
these studies are similar, at least for the first 400-500 
years.  Thereafter, some series indicate especially cold conditions, from 
the late 16th century until the 19th century, but still generally within 
the 2 standard error confidence limits of Mann et al. (1999) (Figure 
6.3composite showing low frequency records from Mann et al, 1999; Jones et 
al. * Crowley & Lowery + Mann et al uncertainties--from Briffa et al 2000 
Fig. 10, without green line[Briffa 2000] and including both sets of 
uncertainty curtains).  The differences between them may be explained by 
the different seasons and spatial coverage of data used.
Each reconstruction represents a somewhat different spatial domain.  In the 
Mann et al. studies, the northern hemisphere mean series is the same 
geographical domain as the gridded instrumental data set available for the 
period 1902-80.  This means that some regions within the northern 
hemisphere (in the central Pacific, central Eurasia and regions beyond 
70N) were not represented.  However, the global eigenvector patterns that 
were reconstructed are based on data from low latitudes and parts of the 
southern hemisphere.  Other reconstructions generally do not include data 
from sub-tropical or tropical regions, and this may explain the colder 
periods in the latter half of the Jones et al. and Crowley and Lowery 
records, if higher latitudes were particularly cold at that time compared 
to the Tropics.
Another reason for the differences in Figure 6.3 may be because each 
reconstruction represents a somewhat different season.  In the Mann et al. 
(1998, 1999) reconstruction, mean annual temperature data were used for 
calibration, since data from both hemispheres were used to constrain the 
eigenvector patterns and data from different regions may have had stronger 
signals in one season than in another.  For example, some data from western 
Europe might contain a strong NAO (winter) signal, whereas data from 
elsewhere might carry a strong summer precipitation signal related to 
ENSO.  Both data sets nevertheless help to define important modes of 
climate anomalies that themselves capture large scale annual temperature 
patterns (Bradley et al., 2000).  Other reconstructions are for summer 
months (April-September) and this may also explain some of the differences 
between the series, for example, if summers were particularly cool in 
extra-tropical regions, in the 17th-19th centuries.
Another critical question in any long-term reconstruction is to what extent 
does the proxy adequately capture the true low frequency nature of the 
climate record.  Given that most of the long-term data used in all of these 
paleotemperature reconstructions are from tree-rings, it is important to 
establish that they are not contaminated by biological growth trends.  This 
matter is especially critical when individual tree ring records, of 
differing record lengths (often limited to a few hundred years) are patched 
together to assess long-term climate changes.  Briffa et al. (2000) have 
carefully evaluated this problem, using a maximum ring width density data 
set that is largely independent of that used by Mann et al. (1998, 
1999).  By combining sets of tree ring density data grouped by the number 
of years since growth began at each site, Briffa et al. provide a 
methodology that largely eliminates the biological growth function 
problem.  They also estimate confidence limits through time (Figure 6.4Fig 
10 from Briffa et al 2000 on its own, 2SE limits).
The Briffa et al series shows similar temperature anomalies as Mann et al. 
in the 15th century (though no sharp decline in temperatures around 
A.D.1450) but markedly colder conditions from A.D. 1500 to ~A.D. 1800 (cf. 
Figures 6.2 and 6.4).  The early 19th century is also colder in the Briffa 
et al. series.  The Mann et al. and Briffa et al. series (Figure 6.3) 
bracket all other paleotemperature estimates for the northern hemisphere, 
such as those by Bradley and Jones (1993), Jones et al. (1998) Briffa et 
al., (1998), Overpeck et al. (1998) and Crowley and Lowery (2000).  The 
Briffa et al reconstruction describes a well-defined minimum in 
temperatures from ~A.D. 1550-1850 that conforms with the consensus view of 
a Little Ice Age (Bradley and Jones,1992).  Though this period was not 
uniformly cold and temperature anomalies differed regionally, overall it 
was significantly below the 1881-1960 mean (by as much as 0.5C for most of 
the 17th century) in the regions studied by Briffa et al. 
(2000).  Independent reconstructions derived from borehole temperatures 
suggest even colder temperatures about 400 to 500 years ago, and/or even 
greater warming in the 20th century.[elaborate]  To what extent these 
differences in reconstructed temperatures are related to the effect of land 
use change on borehole temperatures remains to be resolved, but it suggests 
that land use change may be another factor, in addition to changes in 
atmospheric trace gases and aerosols, that may have to be taken into 
account to realistically simulate past (and future) climate change.
Although the Mann et al. and Briffa et al reconstructions have much in 
common, they are clearly not identical.  One explanation for the 
differences may again lie in the geographical distribution of data used in 
each analysis.  The study of Briffa et al. is strongly weighted towards the 
northern treeline (60-75N) where temperatures were particularly low in the 
17th century; the study by Mann et al. includes data from lower latitudes, 
and incorporates temperature reconstructions from both marine and 
terrestrial regions in calculating a hemispheric mean.  If this is the 
explanation for the differences, it suggests that low latitude regions 
(equatorward of ~35N) did not experience a drop in temperatures in the 
latter half of the last millennium comparable to higher northern 
latitudes.  This further suggests an increase in the Equator-Pole 
temperature gradient during that time.
One thing that all reconstructions clearly agree on is that northern 
hemisphere mean temperature in the 20th century is unique, both in its 
overall average and in the rate of temperature increase.  In particular the 
1990s were exceptionally warm -- probably the warmest decade for at least 
1000 years (even taking the estimated uncertainties of earlier years into 
account).  The last ~50 years also appear to have been the warmest period 
by far (Table 1??).  A caveat to this conclusion is that the current 
proxy-based reconstructions do not extend to the end of the 20th century, 
but are patched on to the instrumental record of the last 2-3 
decades.  This is necessary because many paleo data sets were collected in 
the 1960s and 1970s, and have not been up-dated.  Furthermore, in the case 
of tree rings from some areas (especially at high latitudes) the climatic 
relationships prevalent for most of the century appear to have changed in 
recent decades, possibly because of increasing aridity &/or snowcover 
changes at high latitudes that have altered the ecological responses of 
trees to climate (cf. Jacoby et al; Briffa et al; Vaganov et al., 
1999).  Consequently, it must be recognized that an assessment of the 
unusual nature of the 1990s is necessarily based on a direct comparison of 
instrumental data with long-term proxy-based 
reconstructions.  Nevertheless, the conclusion that temperatures rose at 
unprecedented rates in the 20th century, reaching levels by the end of the 
century that were unprecedented within the last millennium seems to be an 
extremely robust result from these studies. (cf. Pollack et al. 1998??).
Confidence that an accurate reproduction of the recent instrumental record 
would be possible if all the available paleoclimatic data were updated to 
the present is provided by Figure 6.5 (Mann et al. v. instrumental, 
1850-1980).  This shows that a set of proxy data calibrated against the 
1902-1980 period of instrumental data captured mean annual temperatures 
well both during this period and during the preceding 50 years for which an 
independent set of instrumental data is available.  The excellent fit over 
the late 19th century test period provides confidence that an updated set 
of proxy data would also accurately reproduce recent changes.
         Figure 6.3 also shows that the overall range in temperature over 
the last 1000 years has been quite small.  For example, the range in 
50-year means has only been ~0.5C [?cf. both records] (from the coldest 
period in the 15th , 16th  and 19th centuries, to the warmest period of the 
last 50 years: Table 1??).  Within that narrow envelope of variability, all 
of the significant environmental changes associated with the onset and 
demise of the Little Ice Age (~1450-1850) took place.  This puts into 
vivid context the magnitude of projected future changes resulting from 
greenhouse-gas increases and associated feedbacks (Figure 6.6 Mann et al + 
others? + IPCC projections).  Even the low end of model estimates suggest 
additional temperature increases on the order of 1-2C by the end of the 
21st century (Intergovernmental Panel on Climate Change 1996 or 2000?).
The discussion so far has focused exclusively on the northern hemisphere 
record because there are insufficient data currently available to produce a 
very reliable series for the southern hemisphere.  Data from the Mann et 
al. (1998) reconstruction (back to A.D. 1700) averaged for those parts of 
the southern hemisphere that were represented in the instrumental 
calibration period, show a similar temporal pattern to that of the northern 
hemisphere, but generally warmer (less negative anomalies).  However, much 
more work is needed on southern hemisphere proxy records to extend and 
verify that result.

Raymond S. Bradley
Professor and Head of Department
Department of Geosciences
University of Massachusetts
Amherst, MA 01003-5820

Tel: 413-545-2120
Fax: 413-545-1200
Climate System Research Center: 413-545-0659
Climate System Research Center Web Page: 
<http://www.geo.umass.edu/climate/climate.html>
Paleoclimatology Book Web Site (1999): 
http://www.geo.umass.edu/climate/paleo/html



</x-flowed>
