Quaternary Science Reviews 22 (2003) 1701–1716
The temperature of Europe during the Holocene reconstructed
from pollen data
B.A.S. Davisa,b,
*, S. Brewerb
, A.C. Stevensona
, J. Guiotc
, Data Contributors1
a
Department of Geography, University of Newcastle, Newcastle upon Tyne, NE1 7RU, UK
b
IMEP, CNRS UPRES A6116, Facult!e de St J!er#ome, Case 451, 13397 Marseille, Cedex 20, France
c
CEREGE, Europ#ole de l’Arbois, B.P. 80, 13545 Aix-en-Provence, Cedex 04, France
Received 18 December 2002; accepted 22 May 2003
Abstract
We present the first area-average time series reconstructions of warmest month, coldest month and mean annual surface air
temperatures across Europe during the last 12,000 years. These series are based on quantitative pollen climate reconstructions from
over 500 pollen sites assimilated using an innovative four-dimensional gridding procedure. This approach combines threedimensional
spatial gridding with a fourth dimension represented by time, allowing data from irregular time series to be ‘focussed’
onto a regular time step. We provide six regional reconstructed temperature time series as well as summary time series for the whole
of Europe. The results suggest major spatial and seasonal differences in Holocene temperature trends within a remarkably balanced
regional and annual energy budget. The traditional mid-Holocene thermal maximum is observed only over Northern Europe and
principally during the summer. This warming was balanced by a mid-Holocene cooling over Southern Europe, whilst Central
Europe occupied an intermediary position. Changes in annual mean temperatures for Europe as a whole suggest an almost linear
increase in thermal budget up to 7800 BP, followed by stable conditions for the remainder of the Holocene. This early Holocene
warming and later equilibrium has been mainly modulated by increasing winter temperatures in the west, which have continued to
rise at a progressively decreasing rate up to the present day.
r 2003 Elsevier Ltd. All rights reserved.
1. Introduction
A number of attempts have recently been made to
develop dynamic regional and global time series
temperature reconstructions for the last 1000 years
(Mann et al., 1999; Shaopeng et al., 2000; Briffa et al.,
2001). These reconstructions have been used to investigate
the role of various natural and anthropogenic
forcing on the climate system, and the ability of climate
models to reproduce them (Jones et al., 1998). The
ARTICLE IN PRESS
*Corresponding author. Tel.: +44-(0)-191-2226359; fax: +44-(0)-
191-2225421.
E-mail address: basil.davis@ncl.ac.uk (B.A.S. Davis).
1
The data contributors have all provided pollen data for this
study. The subsequent analysis and interpretation is the work and
responsibility of the first four authors. The contributors include:
Allen, J., Almqvist-Jacobson, H., Ammann, B., Andreev, A.A.,
Argant, J., Atanassova, J., Balwierz, Z., Barnosky, C.D., Bartley,
D.D., Beaulieu, JL de, Beckett, S.C., Behre, K.E., Bennett, K.D.,
Berglund, B.E.B., Beug, H-J., Bezusko, L., Binka, K., Birks, H.H.,
Birks, H.J.B., Bj.orck, S., Bliakhartchouk, T., Bogdel I., Bonatti, E.,
Bottema, S., Bozilova, E.D.B., Bradshaw, R., Brown, A.P., Brugiapaglia,
E., Carrion, J., Chernavskaya, M., Clerc, J., Clet, M.,
Co#uteaux, M., Craig, A.J., Cserny, T., Cwynar, L.C., Dambach, K.,
De Valk, E.J., Digerfeldt, G., Diot, M.F., Eastwood, W., Elina, G.,
Filimonova, L., Filipovitch, L., Gaillard-Lemdhal, M.J., Gauthier, A.,
G.oransson, H., Guenet, P., Gunova, V., Hall, V.A.H., Harmata, K.,
Hicks, S., Huckerby, E., Huntley, B., Huttunen, A., Hyv.arinen, H.,
Ilves, E., Jacobson, G.L., Jahns, S., Jankovsk!a, V., J!ohansen, J.,
Kabailiene, M., Kelly, M.G., Khomutova, V.I., K.onigsson, L.K.,
Kremenetski, C., Kremenetskii, K.V., Krisai, I., Krisai, R., Kvavadze,
E., Lamb, H., Lazarova, M.A., Litt, T., Lotter, A.F., Lowe, J.J.,
Magyari, E., Makohonienko, M., Mamakowa, K., Mangerud, J.,
Mariscal, B., Markgraf, V., McKeever, Mitchell, F.J.G., Munuera,
M., Nicol-Pichard, S., Noryskiewicz, B., Odgaard, B.V., Panova,
N.K., Pantaleon-Cano, J., Paus, A.A., Pavel, T., Peglar, S.M.,
Penalba, M.C., Pennington, W., Perez-Obiol, R., Pushenko, M.,
Ralska-Jasiewiszowa, M., Ramfjord, H., Regn!ell, J., Rybnickova, E.,
Rybnickova, M., Saarse, L., Sanchez Gomez, M.F., SarmajaKorjonen,
K., Sarv, A., Seppa, H., Sivertsen, S., Smith, A.G.,
Spiridonova, E.A., Stancikaite, M., Stefanova, J., Stewart, D.A.,
Suc, J-P., Svobodova, H., Szczepanek, K., Tarasov, P., Tobolski, K.,
Tonkov, Sp., Turner, J., Van der Knaap, W.O., Van Leeuwen, J.F.N.,
Vasari, A., Vasari, Y., Verbruggen, C., Vergne, V., Veski, S, Visset, L.,
Vuorela, I., Wacnik, A., Walker, M.J.C., Waller, M.P., Watson, C.S.,
Watts, W.A., Whittington, G., Willis, K.J., Willutzki, H., Yelovicheva,
Ya., Yll, E.I., Zelikson, E.M., Zernitskaya, V.P.
0277-3791/03/$ - see front matter r 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0277-3791(03)00173-2
development of these time series has mainly been based
on annually resolved proxies, particularly tree-rings,
effectively limiting such studies to the last millennia
when annual archives are widely available. On longer
time scales, non-annually resolved proxies such as pollen
data occur more extensively, but the problem of
chronological control has led to the adoption of a
different non-dynamic approach to regional synthesis.
Typically, these have been based on a broad time slice
with samples assimilated within a 500–1000-year time
window around the target time, such as the ‘midHolocene’
60007500 years 14
C BP (COHMAP Members,
1988; Huntley and Prentice, 1993; Cheddadi et al.,
1997). These static map-based reconstructions have been
widely applied to data-model comparisons using climate
models run to equilibrium (Prentice et al., 1997; Masson
et al., 1999).
As computing power continues to increase, then
standard models (AGCMs/CGCMs) can be run for
progressively longer periods. Also, a new type of climate
model has recently been developed called Earth system
models of intermediate complexity (EMICs) (Claussen
et al., 2002) which allow the simulation of climate over
much longer time periods, including the whole Holocene
(Crucifix et al., 2002). These allow the dynamic timedependent
response of the atmosphere to be investigated
against a variety of internal (ice, ocean circulation,
biosphere, trace gases) and external (orbital) forcing
mechanisms (Brovkin et al., 1999; Ganopolski and
Rahmstorf, 2001; Weber, 2001). Evaluation of these
model simulations against actual climate change requires
palaeoclimate data at a comparable temporal and
spatial scale. This requires not only a long-term
(Holocene) time frame and grid-box (continental) scale,
but also a dynamic approach that allows data-model
comparison through time.
In this study, we present an innovative new approach
to non-annual (pollen-based) palaeoclimate data assimilation
and presentation that provides a dynamic and
quantitative view of Continental-scale climate change
compatible with climate model output. This approach
uses a new four-dimensional gridding procedure to
assimilate data from hundreds of sites and thousands of
samples onto a regular spatial grid and time step. We
have applied this method to a palaeo-temperature
dataset derived from pollen samples from sites across
Europe. This dataset was created using an improved
modern-analogue pollen-climate transfer function that
can accommodate non-analogous fossil pollen assemblages.
The reconstructions include seasonal (coldest
month/warmest month) and annual mean temperatures,
providing an all-year perspective on temperature trends.
We present the results as area-average time series
calculated for the whole of Europe and six sub-regions
at a 100-year pseudo-resolution (time step) over the last
12,000 years.
2. Data
2.1. Modern pollen data and climate
The modern pollen surface sample dataset used in the
transfer function consisted of 2363 samples from
throughout North Africa and Europe west of the Urals.
This was based on data from the European Pollen
Database (EPD), the authors, the PANGAEA data
archive, H.J.B. Birks and S. Peglar. All samples were
composed of original raw counts of the full assemblage.
Each sample site was assigned a modern climate based
on interpolation from station data using an artificial
neural network (Guiot et al., 1996).
2.2. Fossil pollen data and age-depth modelling
The fossil pollen dataset consisted of 510 selected
cores from the EPD and PANGAEA data archive,
supplemented by additional data from the authors and
individual contributors (Fig. 1a). All samples from these
cores consisted of original counts based on the full
assemblage. Only cores with absolute dating control
were included (radiocarbon, annual laminations, etc.),
except where an alternative chronology was suggested
by the original author, or where a clear stratigraphic
correlation could be made with an independently dated
adjacent core. Age-depth models were created for each
core based on a calibrated radiocarbon time-scale.
Radiocarbon dates were calibrated using the OXCAL3.5
program (Bronk Ramsey, 2000). All ages quoted are in
calendar years BP (1950). The choice of age-depth model
follows that of the original author where this has been
published or provided with the data. For other cores, the
most appropriate model was fitted (linear, polynomial,
etc.) based on the chronological control points and any
additional published information. The control points
used for each of the cores is shown in Fig. 1c. Age control
for the dataset is based on over 2000radiocarbon dates
and 680 other absolute dates.
3. Methods
3.1. Pollen-climate reconstruction
Fossil pollen samples were assigned a palaeoclimate
using a modern analogue matching technique based on a
training set of modern pollen samples (Guiot, 1990).
This method has been employed in a large number of
studies at both single site (Cheddadi et al., 1998) and
continental (Cheddadi et al., 1997) scales, and is
discussed in detail in many previous papers (e.g. Magny
et al., 2001). In this study we have applied a modification
to the technique, using PFT (Plant Functional
Type) scores (Prentice et al., 1996) to match the fossil
ARTICLE IN PRESS
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–17161702
and modern analogues rather than relying on a limited
range of indicator taxa. We therefore substitute a
modern analogue technique for the neural network
technique used by Peyron et al. (1998) in the first study
to reconstruct climate from PFT scores.
The use of PFT scores increases the number of pollen
taxa that can be used within the analysis, whilst at the
same time reducing the need for taxa-specific modern
analogues. As a result, a much wider range of taxa can
be included, including those that do not occur in the
modern calibration dataset. For instance, Buxus was
once an important forest shrub in the Western
Mediterranean (Yll et al., 1997), but is rarely found
today, and consequently is not included in our training
set. Using the PFT approach we are able to find a
modern analogue for Buxus with other cool-temperate
broad-leaved evergreens such as Hedera and Ilex that
occur more widely today. The prevailing assumption is
that individual taxa that share the same PFT group also
share the same bioclimatic space (Peyron et al., 1998),
which is in turn different from other taxa within other
PFT groups. By grouping taxa in this way, the method is
ARTICLE IN PRESS
Fig. 1. (a) The distribution of pollen sites used in the study. The data for these sites was obtained from the European Pollen Database (), the
PANGAEA database (&), and individual contributors (D). (b) The distribution in time and two-dimensional space of the pollen samples used in the
study. (c) The age-depth control points are shown for each core. These include radiocarbon dates (), together with other absolute dates, such as
those based on Laminations, Uranium series, Argon/Argon and Tephra ( ), as well as the core top where this was contemporary ( Â ). In some cases
these have been supplemented by relative dates based on stratigraphic correlation (J). (d) Following pollen-climate calibration, the data was then
projected onto a 1
grid at a 100-year time step using a four-dimensional interpolation method. A series of statistical tests were then used to define a
sub-set of the gridded data (grey) based on a minimum spatial and temporal sample density. This subset was used in all subsequent analysis.
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–1716 1703
also less sensitive to changes in taxa abundance within
the same PFT group, which may reflect non-climatic
ecological or anthropogenic factors. This therefore
reduces the influence of these factors may have on the
climatic reconstruction.
Pollen taxa were converted into PFT values using the
Peyron et al. (1998) PFT classification scheme for
Europe, which defines 21 PFT groups using 93 pollen
taxa. Evaluation of the method based on leave-one-out
cross-validation using the surface sample dataset is
presented in Table 1. This involves systematically
removing each pollen sample from the training set and
then predicting (reconstructing) its observed climate
using the remaining pollen samples (nÀ1). The results
are reported as the root mean square error (RMSE) and
the coefficient of determination (r2
) (Birks, 1995). The
results show that reconstructed values closely match the
observed values, with a slightly better performance for
the mean temperature of the coldest month (MTCO)
than annual temperature (TANN), with both performing
better than the mean temperature of the warmest
month (MTWA). Following the assignment of a
palaeoclimate and age to each fossil pollen sample, all
the samples were then combined into a single dataset
together with location information (latitude, longitude
and altitude). The distribution of these samples in time
and two-dimensional horizontal space is shown in Fig.
1b, with each core represented by a string of samples
leading back in time.
3.2. Gridding and four-dimensional interpolation
Gridding data allows the changing spatial distribution
of samples over time to be stabilized so that direct
comparisons can be made more easily at the same
locations between different time periods. It also allows
area averages to be calculated which more accurately
reflect the changing conditions across an area than
simple sample averages that may contain bias by virtue
of their distribution. This distribution is 3-D, with sites/
samples located at different altitudes as well as the
horizontal 2-D distribution typically shown on maps.
Each grid point is also located in 3-D space, with gridpoint
altitude calculated from a digital elevation model
(DEM) sampled at the same spatial resolution. Vertical
temperature gradients are often considerably steeper
than horizontal climate gradients, and the correct
representation of topography is important in order to
arrive at an appropriate estimation of an area-average
value. Projection on to a 3-D grid surface can also take
into account changes in vertical temperature gradients
(lapse-rates) through time. These are effectively ignored
in simplistic 2-D maps that assume static vertical
gradients through the use of anomalies (modern
observationÀfossil observation) to account for altitudinal
differences.
Projection of data onto a grid requires the interpolation
of data from the point of observation to the grid
point position. This process becomes less reliable as the
density of the observations decreases and interpolation
distance increases. In order to constrain further the
interpolation process in the estimation of grid point
values and missing observations, Lou et al. (1998) have
proposed that information from observations in space
can be supplemented by observations in time. The
authors found that this temporal–spatial approach to
interpolation provided a more accurate basis for
estimating values in the construction of instrumental
time-series datasets. Palaeoclimate datasets also represent
similar observational time series, although, as we
have already established, the timing of the observation
in many archives and proxies is more uncertain. This
uncertainty, together with that involved in the calibration
process, represents an additional source of noise in
the estimation process. We believe however that detecting
the underlying signal behind this noise can be greatly
aided by applying this spatial–temporal approach within
a large data network. This is because climate has a
strong temporal, as well as spatial component. This is
revealed in the many high- and low-frequency patterns
identified on decadal–centenial (NAO, ENSO), centennial–millennial
(Bond and Dansgaard-Oeschger cycles)
and millennial+(Milankovich cycles) time scales (Wilson
et al., 2000; Bond et al., 2001; Jones et al., 2001).
These patterns have wide regional and even global
footprints, which means their effects will also be felt
over a wide spatial area, and therefore reflected in a
large number of observations.
We have compiled a palaeoclimate dataset from
hundreds of sites and thousands of samples within the
relatively small geographical area of Europe. By treating
this data as a single observational record in 4-D, we
believe we can improve the signal-to-noise ratio to reveal
the underlying pattern of climate change. This is based
on the assumption that while erroneous data will
inevitably occur within the dataset, this will be randomly
distributed within a predominantly reliable set of
observations. The interpolation process acts to isolate
this erroneous data in two ways, first by being guided by
the greater majority of the data, and secondly by
reinforcing the non-random spatial–temporal patterns
ARTICLE IN PRESS
Table 1
Observed and pollen-inferred modern climate values based on leaveone-out
analysis
Climate variable Correlation (r2
) RMSE
MTCO 0.83 2.58
MTWA 0.75 2.25
TANN 0.80 2.35
MTCO; mean temperature of the coldest month, MTWA; mean
temperature of the warmest month, TANN; mean annual temperature.
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–17161704
within the data. Assimilation through interpolation
therefore acts to focus the reconstruction within a haze
of uncertainty.
Interpolation of the data was carried out using a
4-D smoothing spline (Nychka et al., 2000), similar to
the approach used by Lou et al. (1998). This method
uses generalized cross-validation to determine the
optimal fit of the spline volume to the data. Once the
spline model has been built, climatic values were
estimated at the grid points of the study area. The
output grid was one degree by one degree with a
temporal resolution of 100 years. Altitudes were taken
as averages from a 5 min DEM (TerrainBase, Row et al.,
1995). The temporal resolution almost certainly exceeds
the inherent resolution of the data, and should not be
taken to imply that climate events can be resolved within
a 100-year time frame. Rather, it can be compared to
regular plotting of a variable running mean whose time
frame generally exceeds the time interval between the
plotting points. In this case, the chosen interval may or
may not reveal statistically significant events when the
time frame of the running mean approaches that of the
plotting interval. In the case of the present study, we do
not attempt to interpret sub-millennial scale events
shown in the data.
Due to computational limitations, we were unable to
produce a spline model for the entire dataset. Interpolation
was instead carried out for the study area at
each temporal grid point (each 100-year time step). Data
were included within 7300 years of the time period of
interest giving a series of overlapping datasets, each
containing 600 years of samples. The 300-year limits
were chosen by testing the effects of increasing the size
of the time window on the estimated values across the
study area for a number of different periods. Starting
with data at 750 years, the time window was increased
in 50-year steps until no further changes were seen in the
resulting estimations. This limit was found at between
250 and 300 years for all time periods, indicating that
the inclusion of data beyond 300 years either side of the
target time has little significant influence on the result
(Fig. 2a). A data window of 7300 years was therefore
used for each time step, representing a de facto limit of
data independence. This means that any two reconstructions
at time steps 600 years apart have been based on
entirely independent sets of data.
ARTICLE IN PRESS
0.982
0.984
0.986
0.988
0.99
0.992
0.994
0.996
0.998
1
0 100 200 300 400 500
Time window size (years) Time window size (years)
R2(Timej-Timei)
R2(Timej-Timei)
0.982
0.984
0.986
0.988
0.99
0.992
0.994
0.996
0.998
1
0 100 200 300 400 500
-0.2
0
0.2
0.4
0.6
0.8
1
0 500 1000 1500 2000
Lag Distance (km)
MoransIndex
MoransIndex
-0.2
0
0.2
0.4
0.6
0.8
1
0 500 1000 1500 2000
Lag Distance (km)
6000 BP
6000 BP
10000 BP
10000 BP
(a)
(b)
Fig. 2. (a) Comparison of estimations for MTCO for the target time based on an increasing time window (50-year step). Value on y-axis is R2 of
current time window (Timej) compared with previous time window (Timei). Left: 6000 BP; right: 10,000 BP. (b) Moran’s Index calculated for MTCO
over a series of distance lags (in kilometres). Bold line shows calculated index value at each lag; thin lines show limits of expected values calculated for
random permutations of data (99 iterations). Left: 4000 BP; right: 8000 BP.
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–1716 1705
As the quality of the interpolation will vary across the
study area, depending on the distance (in both space and
time) from the samples, a subset of the interpolated grid
points was selected for subsequent analysis. The
selection was made by including only those geographical
grid points that were consistently within a pre-defined
distance of at least one sample over the study period of
12,000 years. The spatial autocorrelation of samples was
studied in several periods to define a reliable spatial
limit. Samples were shown to be positively correlated to
over 1500 km (Fig. 2b), but we chose to apply a more
conservative limit of 500 km. This limit was shown to
have significant positive correlation by Moran’s Index
(po0.01) (Moran, 1950; Cliff and Ord, 1981). The
500 km limit was combined with the previously established
temporal limits of 7300 years to form a minimum
spatial–temporal sample density. Using a GIS, we then
isolated a core area (Fig. 1d) where this minimum
sample density was maintained throughout the study
period.
It has already been mentioned that the technique acts
to isolate the influence of erroneous data. This will of
course only apply when this data represents a minority
of observations within the dataset, a factor influenced by
the quantity of data. To ensure that the reconstructed
climatic signals were based on a large number of sites,
we assimilated the gridded data into six large-scale
regional records. Area-average time series were then
calculated for these records, each based on a large
number of sites.
3.3. Isostatic readjustment
The assignment of altitude to the grid network has
been based on modern topography with no correction
through time for isostatic uplift. The effect of uplift is to
make temperature reconstructions appear warmer than
would be expected during the early Holocene when the
land surface was lower. This could be locally significant
in our NE region where post-glacial uplift has been
calculated at over 300 m in some areas around the Gulf
of Bothnia in the Baltic Sea. Attempts at correcting for
this discrepancy have been made in some studies (e.g.
Rosen et al., 2001), although they are based on the
assumption that lapse rates have remained constant at
around À0.6
C/100 m. These assumptions however are
unlikely to hold true in the lower part of the atmosphere
where lapse rates vary seasonally and spatially due to
interaction with the ground surface. An analysis of lapse
rates using NCEP/NCAR re-analysis data reveals the
current mean annual lapse rate for the NE region in the
upper atmosphere (850–700 mbar) is –0.5
C/100 m.
However, this value falls to –0.3/100 m in the lower
atmosphere (Surface to 850 mb height, or approximately
0–1800 m asl), ranging from 0.0
C/100 m in January to –
0.5
C/100 m in July. These values fluctuated by as much
as 0.3
C in January and 0.1
C in July between 1958 and
98. Other studies have also suggested changes in lapse
rates on Holocene time-scales in temperate (Huntley and
Prentice, 1988), and tropical regions (Peyron et al.,
2000). A large amount of uncertainty therefore remains
over this problem, although the system of 3-D spatial
interpolation corrects for lapse rate change, and the area
most affected by isostatic readjustment remains small
with few core sites. The impact of isostatic changes on
the reconstruction is therefore probably not significant
at the regional scale.
4. Results and discussion
All results are shown as anomalies compared to the 60
BP (1890) reconstruction. The modern À40 BP (1990)
reconstruction was not used as the baseline because this
time step was not based on a balance of samples both
forward and back in time. Reconstructions are represented
by six regional time series (Figs. 3 and 4),
together with summary reconstructions for the whole
European area (Fig. 5). In comparing these results with
data from individual sites or smaller local regions it is
important to remember that they represent area
averages over large regions within which large local
differences can occur. Similarly, the choice of summary
regions has been relatively arbitrary, while the climate
changes shown can be expected to form a continuum
between regions such that selection of a different
geographical region will produce a subtlety different
pattern of change based on the same underlying pattern.
4.1. Northern Europe
Reconstructed summer MTWA anomalies across
Northern Europe during the early Holocene show
values comparable with modern temperatures. These
then rise to a clear maximum around 6000 BP, with the
onset of this rise delayed to around 9000 BP in the east.
At their peak, MTWA anomalies reach +1.5
C in the
NW sector and +1.0
C in the NE sector, before
declining through the remainder of the Holocene.
Winter MTCO anomalies are much greater than late
Holocene values at the start of the Holocene, ranging
from À9.0
C in the NW region, to –2.0
C in the NE.
After a rapid rise at 11,500 BP to around À5.0
C,
temperatures in the NW region then show a steady rise
to present day values. In the NE, temperatures rise
more quickly to values at, or slightly above, late
Holocene values around 7000 BP. Overall annual
TANN anomalies are lower in the NW at the onset of
the Holocene than the NE region, reflecting the lower
winter temperatures. Annual temperatures in both
regions then rise until 6500 BP, before stabilizing in
ARTICLE IN PRESS
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–17161706
the NW, and undergoing a late Holocene decline in the
NE.
A large number of studies have attempted to
reconstruct Holocene temperature changes across
northern Europe from a wide variety of archives and
proxies. Increasingly, these have provided quantitative
estimates at the site scale, although attempts at
systematic regional scale analysis directly comparable
ARTICLE IN PRESS
-4
-3
-2
-1
0
1
2
0 2000 4000 6000 8000 10000 12000
NE
NW
CE
SE
NE
CW
SW
45N
15E
55N
Summer (MTWA)
Winter (MTCO)
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
0 2000 4000 6000 8000 10000 12000
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
0 2000 4000 6000 8000 10000 12000
-6
-5
-4
-3
-2
-1
0
1
2
0 2000 4000 6000 8000 10000 12000
-4
-3
-2
-1
0
1
0 2000 4000 6000 8000 10000 12000
-3
-2
-1
0
1
2
3
0 2000 4000 6000 8000 10000 12000
Year Cal. BP
Year Cal. BP
Year Cal. BP
Year Cal. BP
Year Cal. BPYear Cal. BP
TemperatureAnomaly(C)TemperatureAnomaly(C)TemperatureAnomaly(C)
TemperatureAnomaly(C)TemperatureAnomaly(C)TemperatureAnomaly(C)
NW
SESW
CECW
Fig. 3. Reconstructed area-average summer (MTWA) and winter (MTCO) temperature anomalies for six regions in Europe during the Holocene.
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–1716 1707
with our own estimates remain few. Despite this,
comparison of existing site-specific reconstructions
across Northern Europe indicate close agreement with
our results.
Early Holocene summer temperatures similar to the
present day were also reconstructed in Northern
Scandinavia by Sepp.a and Birks (2001) using pollen
and Ros!en et al. (2001) using a range of proxies,
ARTICLE IN PRESS
-5
-4
-3
-2
-1
0
1
0 2000 4000 6000 8000 10000 12000
TemperatureAnomaly(C)TemperatureAnomaly(C)TemperatureAnomaly(C)
TemperatureAnomaly(C)TemperatureAnomaly(C)TemperatureAnomaly(C)
-3
-2
-1
0
1
2
0 2000 4000 6000 8000 10000 12000
-4
-3
-2
-1
0
1
0 2000 4000 6000 8000 10000 12000
-2
-1
0
1
2
0 2000 4000 6000 8000 10000 12000
-5
-4
-3
-2
-1
0
1
0 2000 4000 6000 8000 10000 12000
-2
-1
0
1
2
3
4
0 2000 4000 6000 8000 10000 12000
Year Cal. BPYear Cal. BP
Year Cal. BPYear Cal. BP
Year Cal. BPYear Cal. BP
NE
NW
CE
SE
NE
CW
SW
45N
15E
55N
Annual (TANN)
NW
SESW
CECW
Fig. 4. Reconstructed area-average mean annual (TANN) temperature anomalies for six regions in Europe during the Holocene.
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–17161708
including pollen, chironomids, diatoms and NIR. In
another quantitative multi-proxy study in Western
Norway, Birks and Ammann (2000) found summer
temperatures at the start of the Holocene increased
quickly to values close to modern levels, and warming
continued into the early Holocene. Other studies
based on macrofossil reconstructions of past tree-lines
have also noted that the early Holocene tree-line
was close to the present day in the Scandes mountains
(Dahl and Nesjke, 1996). These and other studies also
indicate that the early Holocene warming was gradual
(Karl!en, 1998; Lauritzen and Lundberg, 1999; Sepp.a
and Birks, 2001).
Evidence for a mid-Holocene thermal maximum in
Scandinavia is considerable, and based on a wide range
of proxies. Tree-lines reached their maximum altitude
up to 300 m higher than today (Eronen and Zetterberg,
1996; Barnekow and Sandgren, 2001) and glaciers were
much reduced or absent (Karl!en, 1988; Seierstad et al.,
2002). Quantitative reconstructions indicate that temperatures
were up to 2.0
C higher than today (Barnekow,
2000; Barnett et al., 2001; Sepp.a and Birks, 2001,
2002; Ros!en et al., 2001). These mostly refer to summer
temperatures, although Kullman (1995) notes that the
success of pine in the early Holocene in the Scandes
mountains would not have been compatible with colder
(and drier) than normal winters even if the summers
were warmer. Higher temperatures may also have
increased evaporation, contributing to the decline in
mid-Holocene lake levels observed between 8000 and
5800 BP by Hyv.arinen and Alhonen (1994).
Our results suggest the summer thermal maximum
occurred across a wide area of Northern Europe at
around 6000 BP. Evidence from other studies indicate a
range of dates for the timing of the mid-Holocene
thermal maximum, although many of these fall between
6000 and 7000 BP. These include dates of between 7900
and 6700 BP from pollen data (Sepp.a and Birks, 2001),
6200 BP from chironomids (Korhola et al., 2000) and
maximum tree-line altitudes at 6300 BP (Barnekow,
2000) and between 6300 and 4500 BP (Barnekow and
Sandgren 2001; Sepp.a et al., 2002). Land ice cover was
also at a minimum at 6200 BP (Nesje et al., 2001), whilst
glaciers were mainly absent from a catchment in
Western Norway between 9800 and 6700 BP (Seierstad
et al., 2002), and between 7300 and 6100 BP at least the
northern part of the Jostedalsbreen ice cap melted away
(Nesje et al., 2000).
Discrepancies in the timing of the thermal maximum
between studies may be related to local climatic effects,
or edaphic factors in the case of vegetation proxies. The
Scandes mountains form one of the steepest climatic
barriers in Europe, and sites either side of them may be
expected to show contrasting responses. Sepp.a and
Birks (2002) showed that the mid-Holocene maximum
may vary even within a relatively short distance using
the same reconstruction methods and calibration
dataset. For other proxies sensitive to different climatic
ARTICLE IN PRESS
Europe
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
0 2000 4000 6000 8000 10000 12000
Year Cal. BP
TemperatureAnomaly(C)
TemperatureAnomaly(C)
Europe
Summer (MTWA)
Winter (MTCO)
Annual (TANN)
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
0 2000 4000 6000 8000 10000 12000
Year Cal. BP
Europe
Fig. 5. Reconstructed area-average summer (MTWA) and winter (MTCO) temperature anomalies (a), and annual temperature (TANN) anomaly
(b) for Europe during the Holocene.
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–1716 1709
forcing, the contrasting pattern of MTCO and TANN
also suggests that different seasonal sensitivities may
account for inter-proxy discrepancies. For instance, the
timing of the maximum for TANN is 1000 years earlier
(and the anomaly is greater) than MTWA in the NE
sector as a result of warmer spring and/or autumn
temperatures, possibly as a result of more prolonged icefree
conditions in the Barents Sea (Duplessy et al.,
2001). Multi-proxy studies from the same site location
indicate differences in the reconstructed temperature
record between proxies (Ros!en et al., 2001). Korhloa
et al. (2002) also note that the current distribution of
macrofossil evidence used to infer tree-line change may
be as much a result of favourable preservation as
climate.
Following the mid-Holocene maximum, almost all
studies then report a late-Holocene cooling as neoglacial
conditions became established (Ros!en et al., 2001; Sepp.a
and Birks, 2001, 2002; Korhola et al., 2002). Glaciers
expanded (Karl!en, 1988; Nesje et al., 2001; Seierstad
et al., 2002) and tree-lines retreated (Dahl and Nesje,
1996; Barnekow, 2000; Barnett et al., 2001). This pattern
of late-Holocene cooling is repeated in our own data,
more especially in summer temperatures, while winter
temperatures continued to rise in the NW sector.
Temperatures do not however fall below early Holocene
values, and our data does not support the idea of Bigler
et al. (2002) that the last four millennia were the coldest
of the entire Holocene. The largest cooling is in MTWA
in the NW sector and TANN in the NE sector with midHolocene
anomalies of almost +1.5
C. High midHolocene
TANN values in the NE region were also
reconstructed by Shemesh et al. (2001) on the basis of
the d18
O of biogenic silica from Swedish Lapland.
Values however were much higher, and the subsequent
neoglacial cooling estimated at between 2.5 and 4.0
C.
In the NW sector, TANN was balanced between cooling
summer temperatures and warming winter temperatures.
This pattern of relative stability in Holocene
TANN is supported by another d18
O record from a
speleothem in coastal Norway which shows little longterm
trend in TANN after reaching present-day values
around 9000 BP (Lauritzen and Lundberg, 1999).
4.2. Central Europe
Early Holocene summer MTWA anomalies were
lower across Central Europe than Northern Europe,
particularly in the CW sector where anomalies were up
to À2.0
C at the onset of the Holocene. Winter
anomalies were also colder in the CE region than NE
region, although temperature anomalies in the west were
approximately the same following the rapid warming at
the end of the Younger Dryas in the NW region. Both
summer and winter anomalies in the CW region
then follow a similar pattern to the NW region, with a
mid-Holocene summer maximum around 6000 BP,
while winter temperatures continue to rise, with the
overall result that annual temperatures stabilize as
summers cool after 6000 BP. The CE region also shows
many similarities with the NE region, with delayed
summer warming, but also a much less well-defined midHolocene
maximum. This reflects a much lower overall
variation in mid–late-Holocene temperatures in the
Central European area, with no real trend in seasonal
or annual temperatures in the CE region after 8000 BP.
In comparing our results with other palaeoclimate
records from the Central Europe region, it is clear that
far fewer records show the large and coherent Holocene
warming and cooling trends that characterize Northern
Europe. After an initial period of early Holocene
warming, our results show temperature fluctuations
generally within 1.0
C of modern values, in agreement
with Alpine reconstructions (Haas et al., 1998). The
small magnitude of change and lack of clear trend is
probably why many paleoclimate records from the
region show a Holocene climate characterized by
short-term periodic events rather than consistent longterm
trends (e.g. Magny, 1993).
We reconstruct early Holocene summer temperature
anomalies that were cooler across Central Europe than
Northern Europe, whilst MTWA anomalies were still
less than winter MTCO values. Cooler early Holocene
temperatures and colder winter anomalies were also
found by Atkinson et al. (1987) who looked at
Coleoptra evidence from a number of sites in Britain.
In this regional study, MTWA anomalies were consistently
less than MTCO anomalies, whilst overall mean
anomaly values were around 2.0
C cooler than our
reconstruction for the CW sector. Some of this
discrepancy may be accounted for by the fact that a
number of the sites considered in this study lie close to
the NW sector, which experienced greater winter
anomalies during this time.
By 8000 BP, temperatures in all seasons had recovered
across Central Europe to values within 1.0
C of modern
values. In the east, temperatures fluctuated within these
limits for the remainder of the Holocene with no clear
trends. This agrees with a wealth of evidence from
Alpine regions indicating periodic glacier advance and
retreat during the Holocene (Hormes et al., 2001).
Quantitative estimates based on plant macrofossil and
pollen evidence by Haas et al. (1998) also suggest
summer temperature varied within 0.7–0.9
C above
present.
Our results show that the mid-Holocene thermal
maximum at 6000 BP is more clearly defined in the west
than the east, where the warming occurred earlier and
throughout all seasons. In Austria, the Pasterze Glacier
was limited in extent during the early Holocene, and was
smaller than present for an extended period between
8100 and 6900 BP (Nicolussi and Patzelt, 2000). Further
ARTICLE IN PRESS
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–17161710
west, evidence for a later and more pronounced midHolocene
warming is supported by Zoller et al. (1998)
and Haas et al. (1998), who found tree lines were highest
in Switzerland around 6000 BP. Less well-dated studies
have also identified maximum timberline altitudes in the
Alps between 9000 and 4700 BP (Tinner et al., 1996) and
8700 and 5000 BP (Wick and Tinner, 1997). In Italy’s
Aosta Valley, Burga (1991) identified the period 8300–
6000 BP to be the warmest period during the Holocene,
with maximum warmth between 6700 and 6000 BP when
tree-lines were located 100–200 m above present levels.
From this timberline change, Burga (1991) estimated
summer temperatures were 1.5–3.0
C higher than
present during this period. This is higher than our own
estimates, which are more in agreement with the Haas
et al. (1998) study, which is also based on a wider range
of sites.
Following the mid-Holocene maximum, summer
temperatures declined in the west, although winter
temperatures continued to increase. In the Alps,
timberlines retreated, although this has also been partly
attributed to human action (Tinner et al., 1996).
Evidence of cooler conditions is supported by Barber
et al. (1994) who noted changes in peat macrofossils
indicating wetter bog surfaces in NW Europe after 4500
BP, while Mauquoy and Barber (1999) also noted
increasing surface wetness in British bogs in the last
millennium. Circumstantial evidence for cooling also
comes from the recent discovery in the Alps of a
superbly preserved prehistoric man melting from ice
high on the Austrian–Italian border. Radiocarbon
dating of the body suggests that it was buried between
5000 and 5300 BP, immediately preceding neoglaciation
(Baroni and Orombelli, 1996).
Further west in the CW region, a cool early Holocene,
warm mid-Holocene and late-Holocene neoglacial is
also supported by the speleothem oxygen isotope record
from SW Ireland. McDermott et al. (1999) interpreted
low d18
O conditions in the early Holocene as reflecting
cooler conditions, with d18
O increasing between 9000
and 6000 BP as warmer conditions became established,
followed by another cooling trend between 7800 and
3500 BP. However, they also suggested increasing d18
O
since 3500 BP could possibly indicate a return to warmer
conditions. A second higher resolution study (McDermott
et al., 2001) revealed a more complex record, but
overall reflecting the same larger scale changes.
4.3. Southern Europe
The pattern of Holocene temperature change reconstructed
for Southern Europe generally follows a very
different pattern from the regions to the north. Early
Holocene summer MTWA and winter MTCO anomalies
were actually positive over the SE region, and
only slightly negative in the SW in summer. Winter
temperature anomalies were À3.0
C at the Younger
Dryas–Holocene transition over the SW region, but
even here they had recovered close to modern levels by
10,000 BP. Temperatures then fell around 1.5
C across
the whole region and at all seasons up to 8000 BP,
before an almost linear increase up to modern values for
all except winter temperatures in the SE. These did not
fall below present day values at 8000 BP, and have
generally maintained themselves at the same level
through to the present day.
Comparison with existing reconstructions is difficult
for Southern Europe because there have been relatively
few quantitative climate reconstructions, and even
fewer that provide estimates of temperature variables.
Terral and Meng.ual (1999) used the anatomy of
olive charcoal to estimate temperatures during the early
and mid-Holocene in southeast Spain and southern
France. They reconstructed annual temperatures between
1.5
C (France) and 3.5
C (Spain) lower than
present, in agreement with our own reconstructions for
the Western Mediterranean. Similarly, on the basis of
d18
O analysis of a speleothem in Southern France,
McDermott et al. (1999) suggested the climate of
Southern France was cooler than present during the
early to mid-Holocene. Interestingly, the authors noted
that this represented an opposing trend to the
speleothem record from SW Ireland, a comparison also
supported by our own study.
Other palaeoclimate reconstructions suggest changes
in water balance that could have been brought about by
changes in temperature and/or precipitation. Cool
temperatures and/or higher precipitation in the early
to mid-Holocene have been proposed by Harrison and
Digerfeldt (1991) to explain high lake levels throughout
the Mediterranean at this time. They also note that
Holocene aridity was established more abruptly in the
west than the east where lake levels declined more
slowly. This latter finding is compatible with our own
that temperatures (and hence evaporation) during the
important winter recharge period increased along with
summer temperatures in the west, but remained
relatively stable in the east. Other studies also support
high lake levels during the early to mid-Holocene in
locations in both the east (Landmann and Reimer, 1996;
Roberts et al., 2001) and west (Roca and Juli!a, 1997;
Giralt et al., 1999; Reed et al., 2001).
Wetter early to mid-Holocene conditions have also
been suggested from isotopic analysis of fossil charcoal
in southern France (Vernet et al., 1996), as well as
speleothems in Israel (Bar-Matthews et al., 1997).
Further evidence for anomalous cool or wet conditions
comes from the presence of an early Holocene sapropel
in the Mediterranean marine record centred around
8000 BP and spanning between ca 10200 and 6400 BP
(Mercone et al., 2000). The timing of this event coincides
with a reduction in summer MTWA and annual TANN
ARTICLE IN PRESS
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–1716 1711
temperatures in both the SW and SE regions in our own
study. SST reconstructions for this time period remain
ambiguous however, with some authors suggesting
cooler conditions (Kallel et al., 1997; Geraga et al.,
2000), and others warmer conditions (Emeis et al., 2000;
Marchal et al., 2002). In contrast to our own study, the
inferred prevailing climate in a number of marine-based
studies has also invariably been interpreted as warm and
wet (Myers and Rohling, 2000; Rohling and De Rijk,
1999; Ariztegui et al., 2000).
4.4. All of Europe
Assimilating the six regional records gives the total
mean change in area-average temperatures across
Europe (Fig. 5). This reveals the seasonal and annual
thermal budget for the continent, allowing comparison
with global insolation and ice cover changes, which
represent the dominant controls on climate on Holocene
time-scales. The results indicate no long-term trend
through the Holocene in summer MTWA anomalies,
with only a step-wise increase in temperature around
7800 BP. This contrasts with winter MTCO that shows
steadily attenuating anomaly values throughout the
Holocene as temperatures rose to present day levels.
Overall annual temperature change shows that these
seasonal changes occurred within a well-constrained
annual thermal budget, which grew linearly from the
onset of the Holocene up to 7500 BP. After this point,
annual temperatures remained steady for the remainder
of the Holocene.
This reconstruction does not show a mid-Holocene
thermal optimum, as has been suggested by many
authors (Houghton et al., 1990). This is in agreement
with previous pollen-based studies for this period, which
demonstrated that summer warming was confined to
Northern Europe whilst Southern Europe cooled
(Cheddadi et al., 1997). Here we show however that,
not only was high latitude mid-Holocene warming
numerically balanced by low latitude cooling, this
balance was maintained throughout the Holocene. We
can therefore show no summer temperature response at
the European scale to increased summer insolation
(Kutzbach and Webb III, 1993).
The stability of summer temperatures indicates that
the principal control on Holocene temperatures has
come from changes in winter MTCO temperatures.
These show an increase that superficially appears in line
with increasing winter insolation, however, annual
TANN temperatures reveal that this can be more clearly
linked to the decline in residual LGM ice cover that
occurred up to 7500 BP (Kutzbach and Webb III, 1993).
This melting ice also led to rising global sea levels, which
have since stabilized. There is no evidence of a lateHolocene
decline in sea levels that would be expected
with widespread neoglaciation following a mid-Holocene
thermal maximum (Broecker, 1998). Our results are
therefore in agreement with this model of Holocene
climate change.
5. Conclusions
We have shown that by assimilating many thousands
of individual pollen-based proxy-climate observations
through four-dimensions using a GIS, it is possible to
provide an entirely new quantitative and dynamic
perspective on Holocene climate change. The internal
consistency of the results and their agreement with other
proxy records suggests that the influence of local
climatic and non-climatic factors on the reconstruction
method has been limited. This can be attributed to both
the large continental scale of the analysis, and the use of
pollen PFT scores in the calibration process. This has
revealed coherent climatic trends even in Southern
Europe and the Mediterranean, despite a fragmented
and anthropogenically disturbed record. Summer and
winter temperatures show a high degree of independence,
indicating that the reconstruction is not being
restricted by co-variance amongst climate variables.
Trends established in the early to mid-Holocene appear
consistent with those in the later-Holocene, suggesting
that unique vegetation associations found in the early
Holocene (Huntley, 1988) have found valid modern
analogues.
The method shows the importance of maintaining a
balance between the need for careful individual site
interpretation, and a similar need for a large-scale
perspective. Assimilation of sites within a single
database linked through space and time provides the
basis for mutually supportive holistic analysis that is
greater than the sum of the individual sites. This
approach will be greatly improved in future through
the application of probabilistic interpolation techniques
based on full error accounting of the age-depth and
pollen-climate calibration. This in turn will provide a
basis for improving the temporal and spatial resolution
of the reconstructions, and allow the statistical assessment
of their significance.
We have provided in this analysis the first quantitative
assessment of continuously changing seasonal and
annual surface temperatures across Europe during the
Holocene. This analysis has produced a number of
important findings:
1. Significant regional and seasonal variations in
temperature patterns have nevertheless occurred within
a remarkably balanced total energy budget. This budget
has remained stable following the final disappearance of
residual LGM ice around 7800 BP. There has been no
net annual response to seasonal changes in insolation,
and no apparent late Holocene neoglacial cooling at the
European scale.
ARTICLE IN PRESS
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–17161712
2. The traditional mid-Holocene thermal maximum is
shown to be confined to Northern Europe, and more
especially to the summer months. This insolation driven
warming was balanced by a mid-Holocene thermal
minimum over Southern Europe counter to the expected
insolation response. The cooling is also counter to some
marine-based interpretations of mid-Holocene climate
in the Mediterranean.
3. Summer temperature changes have been smaller
than winter temperature changes in all regions apart
from the SE. Changes in winter temperatures have
therefore been a more significant control on the total
energy budget than summer temperatures. The greatest
changes in winter temperatures have been in the
maritime west of Europe where warming has occurred
almost continuously throughout the Holocene.
4. From the mid-Holocene onwards, temperatures in
Central Europe have only shown small-scale changes
without the large-scale warming/cooling trends that
characterise the areas to the north and south. This
stability probably accounts for the preponderance of
studies from this area that argue for a Holocene climate
of short-term fluctuations.
5. Southern Europe and the Mediterranean have
undergone an almost linear warming from around 8000
BP. This warming predates the onset of any major
human impact and continues at the same rate through
the anthropogenically important late-Holocene. This
suggests not only a predominantly natural origin for the
Mediterranean climate, but also that the pollen-climate
calibration method has remained independent of human
impact on the vegetation.
6. Summer MTWA, winter MTCO and annual
TANN temperatures have undergone very different
trends both within and between regions. Attempts to
over-generalize on the basis of seasonally restricted
proxies or geographically restricted archives should be
treated with caution. Only through seasonally and
spatially adjusted area-average calculations can the
effect of seasonal and local scale variations in energy
balances be correctly assessed at annual and continental
scales.
Acknowledgements
The authors would like to acknowledge all those who
have contributed pollen data to this project, and the
facilities offered by the EPD from which the majority of
this data has been accessed. The PANGAEA database
was also utilized in this study, and we would also like to
acknowledge the data and facilities it provides. The
authors would particularly like to thank Rachid
Cheddadi and Jaques-Louis de Beaulieu for their
support and access to the resources of IMEP, John
Birks and Silvia Peglar for additional surface pollen
data, Eniko Magyari for useful feedback and discussions,
Steve Juggins for statistical analysis on the
training set, and to Paddy Mullins for technical
assistance.
Appendix
Data sources: EPD European Pollen Database,
available from: http://www.ngdc.noaa.gov/paleo/
epd.html
PANGAEA Network for Geological and Environmental
Data, available from: http://www.pangaea.de/
NCEP/NCAR Monthly reanalysis data, available from:
http://www.cdc.noaa.gov/cdc/reanalysis/reanalysis.shtml
References
Ariztegui, D., Asioli, A., Lowe, J.J., Trincardi, F., Vigliotti, L.,
Tamburini, F., Chondrogianni, C., Accorsi, C.A., Mazzanti, M.B.,
Mercuri, A.M., Van der Kaars, S., McKenzie, J.A., Oldfield, F.,
2000. Palaeoclimate and the formation of sapropel S1: inferences
from Late Quaternary lacustrine and marine sequences in the
central Mediterranean region. Palaeogeography, Palaeoclimatology,
Palaeoecology 158 (3–4), 215–240.
Atkinson, T.C., Briffa, K.R., Coope, G.R., 1987. Seasonal temperatures
in Britain during the last 22,000 years reconstructed using
beetle remains. Nature 325, 587–592.
Barber, K.E., Chambers, F.M., Maddy, D., Stoneman, R., Brew, J.S.,
1994. A sensitive high resolution record of late Holocene climatic
change from a raised bog in northern England. The Holocene 4,
198–205.
Bar Matthews, M., Ayalon, A., Kaufman, A., 1997. Late Quaternary
paleoclimate in the eastern Mediterranean region from stable
isotope analysis of speleothems at Soreq Cave, Israel. Quaternary
Research 47, 155–168.
Barnekow, L., 2000. Holocene regional and local vegetation history
and lake-level changes in the Tornetrask area, northern Sweden.
Journal of Paleolimnology 23 (4), 399–420.
Barnekow, L., Sandgren, P., 2001. Palaeoclimate and tree-line changes
during the Holocene based on pollen and plant macrofossil records
from six lakes at different altitudes in northern Sweden. Review of
Palaeobotany and Palynology 117, 109–118.
Barnett, C., Dumayne-Peaty, L., Matthews, J.A., 2001. Holocene
climatic change and tree-line response in Leirdalen, central
Jotunheimen, south central Norway. Review of Palaeobotany
and Palynology 117, 119–137.
Baroni, C., Orombelli, G., 1996. The Alpine ‘Iceman’ and Holocene
climatic change. Quaternary Research 46, 78–83.
Bigler, C., Larocque, I., Peglar, S.M., Birks, H.J.B., Hall, R.I., 2002.
Quantitative multiproxy assessment of long-term patterns of
Holocene environmental change from a small lake near Abisko,
northern Sweden. The Holocene 12 (4), 481–496.
Birks, H.J.B., 1995. Quantitative palaeoenvironmental reconstructions.
In: Maddy, D., Brew, J.S. (Eds.), Statistical modeling of
Quaternary Science Data, Technical Guide, Vol. 5. Quaternary
Research Association, Cambridge, pp. 161–254.
Birks, H.H., Ammann, B., 2000. Two terrestrial records of rapid
climatic change during the glacial-Holocene transition (14,000–
9,000 calender years B.P) from Europe. Proceedings of the
National Academy of Sciences 97 (4), 1390–1394.
ARTICLE IN PRESS
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–1716 1713
Bond, G., Bernd, K., Beer, J., Muscheler, R., Evans, M.N., Showers,
W., Hoffman, S., Lotti-Bond, R., Hajdas, I., Bonani, G., 2001.
Persistent solar influence on North Atlantic climate during the
Holocene. Science 294 (5549), 2130–2136.
Briffa, K.R., Osborn, T.J., Schweingruber, F.H., Harris, I.C., Jones,
P.D., Shiyatov, S.G., Vaganov, E.A., 2001. Low-frequency
temperature variations from a northern tree ring density network.
Journal of Geophysical Research 106 (d3), 2929–2941.
Broecker, W.S., 1998. The end of the present interglacial: how and
when? Quaternary Science Reviews 17, 689–694.
Bronk Ramsey, C., 2000. Radiocarbon calibration program available
from: http://www.rlaha.ox.ac.uk/orau/06 01.htm.
Brovkin, V., Ganopolski, A., Claussen, M., Kubatzki, C., Petoukhov,
V., 1999. Modelling climate response to historical land cover
change. Global Ecology and Biogeography 8 (6), 509–517.
Burga, C.A., 1991. Vegetation history and paleoclimatology of the
middle Holocene—pollen analysis of Alpine peat bog sediments,
covered formerly by the Rutor Glacier, 2510 m Aosta Valley, Italy.
Global Ecology and Biogeography Letters 1 (5), 143–150.
Cheddadi, R., Yu, G., Guiot, J., Harrison, S.P., Prentice, I.C., 1997.
The climate of Europe 6000 years ago. Climate Dynamics 13, 1–9.
Cheddadi, R., Lamb, H.F., Guiot, J., van der Kaars, S., 1998.
Holocene climatic change in Morocco: a quantitative reconstruction
from pollen data. Climate Dynamics 14, 883–890.
Claussen, M., Mysak, L.A., Weaver, A.J., Crucifix, M., Fichefet, T.,
Loutre, M.F., Weber, S.L., Alcamo, J., Alexeev, V.A., Berger, A.,
Calov, R., Ganopolski, A., Goose, H., Lohmann, G., Lunkeit, F.,
Mokhov, I.I., 2002. Earth system models of intermediate complexity:
closing the gap in the spectrum of climate system models.
Climate Dynamics 18 (7), 579–586.
Cliff, A.D., Ord, J.K., 1981. Spatial Processes: Models and Applications.
Pion, London, 266pp.
COHMAP members, 1988. Climatic changes of the last 18,000 years.
Observations and model simulations. Science 241, 1043–1052.
Crucifix, M., Loutre, M.F., Tulkens, P., Fichefet, T., Berger, A., 2002.
Climate evolution during the Holocene: a study with an Earth
System model of intermediate complexity. Climate Dynamics 19
(1), 43–60.
Dahl, S.O., Nesje, A., 1996. A new approach to calculating Holocene
winter precipitation by combining glacier equilibrium line altitudes
and pine-tree limits: a case study from Hardangerj^kulen, central
south Norway. The Holocene 6, 381–398.
Duplessy, J.C., Ivanova, E., Murdmaa, I., Paterne, M., Labeyrie, L.,
2001. Holocene paleoceanography of the northern Barents Sea and
variations of the northward heat transport by the Atlantic Ocean.
Boreas 30 (1), 2–16.
Emeis, K.-C., Struck, U., Schulz, H.-M., Rosenberg, R., Bernasconi,
S., Erlenkeuser, H., Sakamoto, T., Martinez-Ruiz, F., 2000.
Temperature and salinity variations of Mediterranean Sea surface
waters over the last 16,000 years from records of planktonic stable
oxygen isotopes and alkenone unsaturation ratios. Palaeogeography,
Palaeoclimatology, Palaeoecology 158, 259–280.
Eronen, M., Zetterberg, P., 1996. Climatic changes in Northern
Europe since late glacial times, with special reference to dendroclimatological
studies in northern Finnish Lapland. Geophysica 2,
35–60.
Ganopolski, A., Rahmstorf, S., 2001. Rapid changes of glacial climate
simulated in a coupled climate model. Nature 409 (6817), 153–158.
Geraga, M., Tsaila-Monopolis, St., Ioakim, Chr., Papatheodorou, G.,
Ferentinos, G., 2000. Evaluation of palaeoenvironmental changes
during the last 18,000 years in the Myrtoon Basin, SW Aegean Sea.
Palaeogeography, Palaeoclimatology, Palaeoecology 156, 1–17.
Giralt, S., Burjachs, F., Roca, J.R., Juli!a, R., 1999. Late Glacial to
Early Holocene environmental adjustment in the Mediterranean
semi-arid zone of the Salinas playa-lake Alicante, Spain. Journal of
Paleolimnology 21, 449–460.
Guiot, J., 1990. Methodology of paleoclimate reconstruction from
pollen in France. Palaeogeography, Palaeoclimatology, Palaeoecology
80, 49–69.
Guiot, J., Cheddadi, R., Prentice, I.C., Jolly, D., 1996. A method of
biome and land surface mapping from pollen data: application to
Europe 6000 years ago. Palaeoclimates 1, 311–324.
Haas, J.N., Rischoz, I., Tinner, W., Wick, L., 1998. Synchronous
Holocene climatic oscillations recorded on the Swiss Plateau and at
the timberline in the Alps. The Holocene 8 (3), 301–309.
Harrison, S.P., Digerfeldt, G., 1991. European lakes as palaeohydrological
and palaeoclimatic indicators. Quaternary Science Reviews
12, 233–248.
Hormes, A., M.uller, B.U., Schl.uchter, C., 2001. The Alps with little
ice: evidence for eight Holocene phases of reduced glacier extent in
the Central Swiss Alps. The Holocene 11 (3), 255–265.
Houghton, J.T., Jenkins, G.J., Ephraums, J.J. (Eds.), 1990. Climate
Change: the IPCC Scientific Assessment. Cambridge University
Press, Cambridge, UK.
Huntley, B., 1988. In: Huntley, B., Webb III, T. (Eds.),
Vegetation History. Glacial and Holocene vegetation history:
Europe. Kluwer Academic Publishers, Dordrecht, ISBN 90-6193-
188-6, 341–383.
Huntley, B., Prentice, I.C., 1988. July temperatures in Europe from
pollen data, 6000 years before present. Science 241, 687–690.
Huntley, B., Prentice, I.C., 1993. In: Wright, H.E., Kutzbach, J.E.,
Webb III, T., Ruddiman, W.F., Street-Perrott, F.A., Bartlein, P.J.
(Eds.), Global Climates since the Last Glacial Maximum. Holocene
Vegetation and Climates of Europe. University of Minnesota Press,
Minnesota, MN, pp. 136–168.
Hyv.arinen, H., Alhonen, P., 1994. Holocene lake-level changes in the
Fennoscandian tree-line region, western Finish Lapland: diatom
and cladoceran evidence. The Holocene 4, 251–258.
Jones, P.D., Briffa, K.R., Barnett, T.P., Tett, S.F.B., 1998. Highresolution
palaeoclimatic records for the last millennium: interpretation,
integration and comparison with General Circulation
Model control-run temperatures. The Holocene 8 (4), 455–471.
Jones, P.D., Osborn, T.J., Briffa, K.R., 2001. The evolution of climate
over the last millennium. Science 292 (5517), 662–667.
Kallel, N., Paterne, M., Labeyrie, L., Duplessy, J.-C., Arnold, M.,
1997. Temperature and salinity records of the Tyrrhenian Sea
during the last 18,000 years. Palaeogeography, Palaeoclimatology,
Palaeoecology 135, 97–108.
Karl!en, W., 1988. Scandinavian glacial climatic fluctuations during the
Holocene. Quaternary Science Reviews 7, 199–209.
Karl!en, W., 1998. Climate variations and the enhanced greenhouse
effect. Ambio 27, 270–274.
Korhola, A., Weckstrom, J., Holmstrom, L., Erasto, P., 2000. A
quantitative Holocene climatic record from diatoms in northern
Fennoscandia. Quaternary Research 54 (2), 284–294.
Korhola, A., Vasko, K., Toivonen, H.T.T., Olander, H., 2002.
Holocene temperature changes in northern Fennoscandia reconstructed
from chironomids using Bayesian modelling. Quaternary
Science Reviews 21 (16–17), 1841–1860.
Kullman, L., 1995. Holocene tree-limit and climate history from the
Scandes Mountains, Sweden. Ecology 768, 2490–2502.
Kutzbach, J.E., Webb III, T., 1993. In: Wright, H.E., Kutzbach, J.E.,
Webb III, T., Ruddiman, W.F., Street-Perrott, F.A., Bartlein, P.J.
(Eds.), Global Climates since the Last Glacial Maximum.
Conceptual Basis for Understanding Late-Quaternary Climates.
University of Minnesota Press, Minnesota, MN, pp. 5–11.
Landmann, G., Reimer, A., 1996. Climatically induced lake level
changes at Lake Van, Turkey, during the Pleistocene/Holocene
transition. Global Biogeochemical Cycles 10 (4), 797–808.
Lauritzen, S.-E., Lundberg, J., 1999. Calibration of the speleothem
delta function: an absolute temperature record for the Holocene in
northern Norway. The Holocene 9 (6), 659–669.
ARTICLE IN PRESS
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–17161714
Lou, Z., Wahba, G., Johnson, D.R., 1998. Spatial–temporal analysis
of temperature using smoothing spline ANOVA. Journal of
Climate 11 (1), 18–28.
Magny, M., 1993. Holocene fluctuations of lake levels in the French
Jura and sub-Alpine ranges, and their implications for past general
circulation patterns. The Holocene 3 (4), 306–313.
Magny, M., Guiot, J., Schoellammer, P., 2001. Quantitative reconstruction
of Younger Dryas to mid-Holocene palaeoclimates at Le
Locle, Swiss Jura, using pollen and lake-level data. Quaternary
Research 56, 170–180.
Mann, M.E., Bradley, R.S., Hughes, M.K., 1999. Northern Hemisphere
temperatures during the last millennium: inferences,
uncertainties and limitations. Geophysical Research Letters 26,
759–762.
Marchal, O., Cacho, I., Stocker, T., Grimalt, J.O., Calvo, E., Martrat,
B., Shackleton, N., Vautravers, M., Cortijo, E., van Kreveld, S.,
Andersson, C., Ko@, N., Chapman, M., Sbaffi, L., Duplessy, J.-C.,
Sarnthein, M., Turon, J.-L., Duprat, J., Jasen, E., 2002. Apparent
long-term cooling of the sea surface in the northeast Atlantic and
Mediterranean during the Holocene. Quaternary Science Reviews
21, 455–483.
Masson, V., Cheddadi, R., Braconnot, P., Joussaume, S., Texier, D.,
PMIP participants., 1999. Mid-Holocene climate in Europe: what
can we infer from PMIP model-data comparison? Climate
Dynamics 15, 163–182.
Mauquoy, D., Barber, K., 1999. Evidence for climatic deteriorations
associated with the decline of Sphagnum imbricatum Horsch. ex
Russ. in six ombrotrophic mires from northern England and the
Scottish Borders. The Holocene 9 (4), 423–437.
McDermott, F., Frisia, S., Huang, Y., Longinelli, A., Spiro, B.,
Heaton, T.H.E., Hawkesworth, C.J., Borsato, A., Keppens, E.,
Fairchild, I.J., van der Borg, K., Verheyden, S., Selmo, E., 1999.
Holocene climate variability in Europe: evidence from d18
O,
textural and extension-rate variations in three speleothems.
Quaternary Science Reviews 18, 1021–1038.
McDermott, F., Mattey, D.P., Hawkesworth, C., 2001.
Centennial-scale Holocene climate variability revealed by
a high-resolution Speleothem d18
O record from SW Ireland.
Science 294, 1328–1331.
Mercone, D., Thompson, J., Croudace, I.W., Siani, G., Paterne, M.,
Troelstra, S., 2000. Duration of S1, the most recent sapropel in the
eastern Mediterranean Sea, as indicated by accelerator mass
spectrometry radiocarbon and geochemical evidence. Paleoceanography
15 (3), 336–347.
Moran, P.A.P., 1950. Notes on continuous stochastic phenomena.
Biometrika 37, 17–23.
Myers, P.G., Rohling, E.J., 2000. Modeling a 200-yr interruption of
the Holocene Sapropel S-1. Quaternary Research 53 (1), 98–104.
Nesje, A., Olaf Dahl, S., Andersson, C., Matthews, J.A., 2000. The
lacustrine sedimentary sequence in Sygneskardvatnet, western
Norway: a continuous, high-resolution record of the Josteldalsbreen
ice cap during the Holocene. Quaternary Science Reviews 19,
1047–1065.
Nesje, A., Matthews, J.A., Olaf Dahl, S., Berrisford, M.S., Andersson,
C., 2001. Holocene glacier fluctuations of Flatebreen and winterprecipitation
changes in the Jostedalsbreen region, western Norway,
based on glaciolacustrine sediment records. The Holocene 11
(3), 267–280.
Nicolussi, K., Patzelt, G., 2000. Discovery of early-Holocene wood
and peat on the forefield of the Pasterze Glacier, Eastern Alps,
Austria. The Holocene 10 (2), 191–199.
Nychka, D., Meiring, W., Royle, J.A., Fuentes, M., Gilleland, E.,
2000. FIELDS: S tools for spatial data. Available at http://
www.cgd.ucar.edu/stats/software.
Peyron, O., Guiot, J., Cheddadi, R., Tarasov, P., Reille, R., de
Beaulieu, J.-L., Bottema, S., Andrieu, V., 1998. Climatic reconstruction
in Europe for 18,000 yr BP from pollen data. Quaternary
Research 49, 183–196.
Peyron, O., Jolly, D., Bonnefille, R., Vincens, A., Guiot, J., 2000.
Climate of East Africa 6000 14
C Yr B.P. as inferred from pollen
data. Quaternary Research 54, 90–101.
Prentice, I.C., Guiot, J., Huntley, B., Jolly, D., Cheddadi, R., 1996.
Reconstructing biomes from palaeoecological data: a general
method and its application to European pollen data at 0 and 6
ka. Climate Dynamics 12, 185–194.
Prentice, I.C., Harrison, S.P., Jolly, D., Guiot, J., 1997. The climate
and biomes of Europe at 6000 yr bp: comparison with model
simulations and pollen-based reconstructions. Quaternary Science
Reviews 17 (6–7), 659–668.
Reed, J.M., Stevenson, A.C., Juggins, S., 2001. A multi-proxy record
of Holocene climatic change in southwestern Spain: the Laguna
Medina, Cadiz. Holocene 11 (6), 707–719.
Roberts, N., Reed, J.M., Leng, M.J., Kuzucuoglu, C., Fontugne, M.,
Bertaux, J., Woldring, H., Bottema, S., Black, S., Hunt, E.,
Karabiyikoglu, M., 2001. The tempo of Holocene climatic change
in the eastern Mediterranean region: new high-resolution craterlake
sediment data from central Turkey. Holocene 11 (6), 721–736.
Roca, J.R., Juli!a, R., 1997. Late-Glacial and Holocene lacustrine
evolution based on Ostracode assemblages in Southeastern Spain.
Giobios 30 (6), 823–830.
Rohling, E.J., De Rijk, S., 1999. Holocene climate optimum and Last
Glacial Maximum in the Mediterranean: the marine oxygen isotope
record. Marine Geology 153, 57–75.
Ros!en, P., Segerstr.om, U., Eriksson, L., Renberg, I., Birks, H.J.B.,
2001. Holocene climatic change reconstructed from diatoms,
chironomids, pollen and near-infrared spectroscopy at an alpine
lake (Sjuodjijaure) in northern Sweden. The Holocene 11 (5),
551–562.
Row III, L.W., Hastings, D., Dunbar, P., 1995. Terrain Base,
Worldwide Digital Terrain Data, CD-ROM and Documentation
Manual. NOAA, National Geophysical Data Center, Boulder,
Colarado.
Seierstad, J., Nesje, A., Dahl, S.O., Simonsen, J.R., 2002. Holocene
glacier fluctuations of Grovabreen and Holocene snow-avalanche
activity reconstructed from lake sediments in Groningstolsvatnet,
western Norway. The Holocene 12 (2), 211–222.
Sepp.a, H., Birks, H.J.B., 2001. July mean temperature and annual
precipitation trends during the Holocene in the Fennoscandian
tree-line area: pollen-based climate reconstructions. The Holocene
11 (5), 527–539.
Sepp.a, H., Birks, H.J.B., 2002. Holocene climate reconstructions from
the fennoscandian tree-line area based on pollen data from
Toskaijavri. Quaternary Research 572, 191–199.
Sepp.a, H., Nyman, M., Korhola, A., Weckstrom, J., 2002. Changes in
tree-lines and alpine vegetation in relation to post-glacial climate
dynamics in northern Fennoscandia based on pollen and chironomid
records. Journal of Quaternary Science 17 (4), 287–301.
Shaopeng, H., Pollack, H.N., Po-Yu, S., 2000. Temperature trends
over the past five centuries reconstructed from borehole temperatures.
Nature 403(6771), 756–758.
Shemesh, A., Rosqvist, G., Rietta-Shati, M., Rubensdotter, L., Bigler,
C., Yam, R., Karl!en, W., 2001. Holocene climatic change in
Swedish Lapland inferred from an oxygen-isotope record of
lacustrine biogenic silica. The Holocene 11 (4), 447–454.
Terral, J.-F., Meng.ual, X., 1999. Reconstruction of holocene climate in
southern France and eastern Spain using quantitative anatomy of
olive wood and archaeological charcoal. Palaeogeography, Palaeoclimatology,
Palaeoecology 153, 71–92.
Tinner, W., Ammann, B., Germann, P., 1996. Treeline fluctuations
recorded for 12,500 years by soil profiles, pollen and plant
macrofossils in the Central Swiss Alps. Arctic and Alpine Research
28 (2), 131–147.
ARTICLE IN PRESS
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–1716 1715
Vernet, J.-L., Pachiaudi, C., Bazile, F., Durand, A., Fabre, L., Heinz,
C., Solari, M.-E., Thi!ebault, S., 1996. Le d13
C de charbons de bois
pr!ehistoriques et historiques m!editerran!eens, de 35000 BP "a l’actuel.
Premier r!esultats. C. R. Acad. Sci. Paris S!er. IIs a t. 323, 319–324.
Weber, S.L., 2001. The impact of orbital forcing on the climate of an
intermediate-complexity coupled model. Global and Planetary
Change 30, 7–12.
Wick, L., Tinner, W., 1997. Vegetation changes and timberline
fluctuations in the central airs as indicators of Holocene climatic
oscillations. Arctic and Alpine Research 29 (4), 445–458.
Wilson, R.C.L., Drury, S.A., Chapman, J.L., 2000. The great ice age.
Routledge, ISBN 0-415-19841-0, 267pp.
Yll, E.-R., Perez-Obiol, R., Pantaleon-Cano, J., Roure, J.M., 1997.
Palynological evidence for climatic change and human activity
during the Holocene on Minorca Balearic Islands. Quaternary
Research 48, 339–347.
Zoller, H., Athanasiadis, N., Heitz-Weniger, A., 1998. Late-glacial and
Holocene vegetation and climate change at the Palu glacier,
Bernina pass, Grisons canton, Switzerland. Vegetation History and
Archaeobotany 7 (4), 241–249.
ARTICLE IN PRESS
B.A.S. Davis et al. / Quaternary Science Reviews 22 (2003) 1701–17161716