Atmospheric circulation variability in the North-Atlantic-European area since the mid-seventeenth century

Abstract

Based on monthly mean sea level pressure grids objectively reconstructed by Luterbacher et al. variations of dynamical modes of the atmospheric circulation for January and July are described by novel indices for running 31-year periods between 1659 and 1999. These indices reflect the continuous evolution of the atmospheric circulation not only with regard to frequency changes of major dynamical modes but also in terms of internal changes within each mode concerning both dynamic (vorticity, intensity) and climatic properties (Central European temperature and precipitation during occurrence of each mode, respectively). Results indicate the great importance of within-mode variations: the zonal circulation mode in January, varying in frequency with long-term cycles, primarily changed its dynamic and climatic properties (towards higher indices) during the transition from the Little Ice Age to modern conditions between 1800 and 1930. Within the Russian High mode of January a change in preference from easterly to westerly patterns above Central Europe occurred around 1850. For July, a striking frequency maximum of the westerly mode at the end of the eighteenth century coincided with a period of marked summer warmth in Central Europe due to negative/positive deviations in vorticity/temperature during occurrence of this mode. The long-term evolution in July indicates a general increase of anticyclonic conditions strengthening during the last 50 years towards a unique phenomenon within the last centuries. The strong increase in the winter-time westerly circulation during the last decades, however, does not appear extraordinary in view of the low-frequency variations of this mode.

Full text

Page 1

J. Jacobeit Æ H. Wanner Æ J. Luterbacher Æ C. Beck
A. Philipp Æ K. Sturm
Atmospheric circulation variability in the North-Atlantic-European area
since the mid-seventeenth century
Received: 7 December 2001 /Accepted: 16 July 2002 / Published online: 26 September 2002
? Springer-Verlag 2002
Abstract Based on monthly mean sea level pressure
grids objectively reconstructed by Luterbacher et al.
variations of dynamical modes of the atmospheric cir-
culation for January and July are described by novel
indices for running 31-year periods between 1659 and
1999. These indices reflect the continuous evolution of
the atmospheric circulation not only with regard to
frequency changes of major dynamical modes but also in
terms of internal changes within each mode concerning
both dynamic (vorticity, intensity) and climatic proper-
ties (Central European temperature and precipitation
during occurrence of each mode, respectively). Results
indicate the great importance of within-mode variations:
the zonal circulation mode in January, varying in fre-
quency with long-term cycles, primarily changed its
dynamic and climatic properties (towards higher indices)
during the transition from the Little Ice Age to modern
conditions between 1800 and 1930. Within the Russian
High mode of January a change in preference from
easterly to westerly patterns above Central Europe oc-
curred around 1850. For July, a striking frequency
maximum of the westerly mode at the end of the eigh-
teenth century coincided with a period of marked sum-
mer warmth in Central Europe due to negative/positive
deviations in vorticity/temperature during occurrence of
this mode. The long-term evolution in July indicates a
general increase of anticyclonic conditions strengthening
during the last 50 years towards a unique phenomenon
within the last centuries. The strong increase in the
winter-time westerly circulation during the last decades,
however, does not appear extraordinary in view of the
low-frequency variations of this mode.
1 Introduction
Most of the empirical investigations on atmospheric
circulation variability are confined to periods not ex-
ceeding some 15 decades at best due to the restricted
availability of appropriate observations. For longer time
periods various proxy data have been used, for example
from ice cores (Appenzeller et al. 1998), tree rings (Cook
et al. 1998), both in combination (Cullen et al. 2000;
Cook et al. 2002; Glueck and Stockton 2001) or from
documentary sources (Garcia et al. 2000; Rodrigo et al.
2001), to reconstruct particular aspects of circulation
dynamics. These studies, however, are constrained to
refer to more simple indices describing circulation vari-
ability (e.g. indices for the North Atlantic Oscillation
NAO) rather than allowing analyses of the variations in
large-scale fields of pressure or other dynamic parame-
ters. In addition, objections have recently been raised
concerning the reliability of proxy based indices to
represent circulation variability sufficiently (Schmutz
et al. 2000; Cullen et al. 2000).
Referring to the North-Atlantic-European region,
large-scale pressure fields for historical periods were first
derived subjectively by Lamb and Johnson (1966) back
to the 1750s and, on a daily scale, by Kington (1988) for
the 1780s. Later on, monthly SLP maps were recon-
structed for the Late Maunder Minimum period
(Wanner et al. 1994) and for anomalous months of the
sixteenth century (Jacobeit et al. 1999). Monthly mean
sea-level pressure (SLP) grids objectively reconstructed
from observed historical time series by means of EOF-
regression techniques were first produced by Jones et al.
(1987) going back to 1780 for the North-Atlantic-
European area. During the 1990s observed historical
time series became increasingly available in particular
Climate Dynamics (2003) 20: 341–352
DOI 10.1007/s00382-002-0278-0
J. Jacobeit (&) Æ C. Beck Æ A. Philipp Æ K. Sturm
Institute of Geography, University of Wu ¨ rzburg,
Am Hubland, D-97074 Wu ¨ rzburg, Germany
E-mail: jucundus.jacobeit@mail.uni-wuerzburg.de
H. Wanner Æ J. Luterbacher
Institute of Geography, University of Berne,
Hallerstr. 12, CH-3012 Bern, and
National Center of Competence in Research (NCCR) in Climate,
Erlachstr. 9a, 3012 Bern, Switzerland

Page 2

for many regions of Europe (e.g. Jones et al. 1997;
Ba ¨ rring et al. 1999). Thus, enhanced or further extended
reconstructions of historical SLP grids have been per-
formed by Jones et al. (1999) and Luterbacher et al.
(2002a) with the latter one extending back to 1659 on a
monthly scale and even back to 1500 on a seasonal
scale.
These objectively reconstructed grids now allow a
considerable extension of periods to be statistically
analysed for circulation variability. Such extended pe-
riods are highly important for evaluating circulation
types, amplitudes and sequences of natural variability
which are only partly included within short intervals
such as the NCEP/NCAR reanalysis-period of just half
a century (Kalnay et al. 1996; Kistler et al. 2001). Fur-
thermore, the extended periods reach back into times
without significant man-made impacts on the climate
system thus reflecting natural variability without the
superimposed human forcings of more recent times.
Improved knowledge of natural variability is a key
factor for understanding the dynamics of man-made
climate changes in particular.
Objective circulation analyses going further back in
time than the last 150 years and based on objective
pressure reconstructions have only been performed
during the last few years. At first, studies on the NAO
variability were increasingly extended into the historical
past, back to 1864 by Hurrell (1995), to 1821 by Jones
et al. (1997), to 1780 by Jacobeit et al. (1998), to 1675
and finally to 1500 by Luterbacher et al. (1999, 2002b).
Further circulation indices based on pressure differences
were calculated, e.g. the EU index (Luterbacher et al.
1999) or zonal indices for Western (Slonosky et al. 2000)
and Central Europe (Jacobeit et al. 2001a). Besides such
simple indices analyses were increasingly based on the
reconstructed large-scale pressure fields themselves, thus
obtaining more sophisticated NAO indices from PCA
studies (Pozo-Va ´ zquez et al. 2000; Portis et al. 2001) and
large-scale circulation patterns for the whole North-
Atlantic-European area (see Wanner et al. 1995 and
Luterbacher et al. 2001, referring to the Late Maunder
Minimum period). Long-term variability of European
pressure fields was first investigated by Schmutz and
Wanner (1998) applying a correlation-based classifica-
tion to the former version of the reconstructed SLP grids
from Jones et al. (1987). Referring to the enhanced re-
constructions back to 1780 (Jones et al. 1999) further
classifications based on EOF-clustering techniques and
an objective Grosswettertyping were carried out by Beck
(2000). Variability of EOF and PCA derived circulation
patterns for the same period is discussed in several pa-
pers: Slonosky et al. (2000) produced patterns from the
monthly SLP station series of the entire year, Jacobeit
et al. (2001b) generated patterns for January and July
from the reconstructed SLP grids. The latter study also
looked at changes of these patterns themselves by
comparing SLP composites between particular periods.
Beck et al. (2001) additionally showed changes in
Central European temperature and precipitation for
subsamples defined by the occurrence of a particular
circulation pattern, respectively.
In contrast to these studies this paper not only will
extend the analysed period by using the most recent re-
constructions back to the mid-17th century (Luterbacher
et al. 2002a), but will also proceed in the consideration of
dynamic aspects important for historical analyses. Thus,
circulation variability will be examined not only with
regard to frequency changes of major dynamical modes
but also in terms of internal changes within each mode
concerning both dynamic (vorticity, intensity) and cli-
matic properties (temperature/precipitation during oc-
currence of each mode, respectively). All these aspects
have to be taken into account for a better understanding
of natural climate variability. Appropriate indices of
circulation variability required for this extended analysis
will be defined and discussed in the methods section.
2 Data
Circulation analyses are based on the monthly mean SLP fields
objectively reconstructed by Luterbacher et al. (2002a) for the
North-Atlantic-European area from 30?N to 70?N and from 30?W
to 40?E. These monthly reconstructions extend back to the year
1659, further back (to 1500) only seasonally resolved SLP fields are
available so far. Here we use an updated release with a 1? spatial
resolution, i.e. with 2911 grid points covering the study area.
The SLP grids have been reconstructed by means of an EOF-
regression technique with predictors comprising all available early
instrumental time series as well as several climatic indices based on
documentary proxy data (Pfister 1999; Glaser 2001) from various
sites of Europe (see Luterbacher et al. 2002a for details and spec-
ifications). These indices have to be included since early instru-
mental pressure time series are sparsely distributed and, except for
Paris and London (Slonosky et al. 2001b), do not start before the
eighteenth century (1722 at Uppsala as the earliest date). As soon
as station pressure predictors become available, these are by far the
most important for SLP reconstructions explaining much more
pressure variability than precipitation or temperature time series
(Luterbacher et al. 2002a). But since temperature and precipitation
have been involved in pressure reconstructions, SLP variability
must not be used, to exclude circular arguments, for explaining
historical climate variability. Indices of temperature and precipi-
tation, however, may well be used for climatic characterisations of
each circulation mode (see next section). This will be done by
means of gridded data for temperature and precipitation similarly
reconstructed for the continental areas of Europe (Luterbacher
et al. submitted 2002). Climatic grids have been reconstructed only
from temperature and precipitation predictors and will be available
with a 0.5? resolution according to the CRU grids from the twen-
tieth century (New et al. 2000) used for model calibration.
Reconstructive skill for all these grids spatially and temporally
varies depending on availability and quality of predictive data: in
general, it increases with time, is better for winter than for summer,
reaches its maximum around Central Europe and declines towards
peripherical regions of the study area (Luterbacher et al. 2002a).
From the mid-seventeenth century onwards, implying monthly grid
resolutions, reconstructive skill should be sufficient for grids to be
submitted to variability analyses on monthly to decadal time scales.
3 Methods
Studies of atmospheric circulation variability not only in terms of
locally restricted indices but rather with respect to complete large-
scale pressure fields, require either well-defined categories of
342Jacobeit et al.: Atmospheric circulation variability in the North-Atlantic-European area

Page 3

atmospheric states or dynamical modes of variability to have been
derived. The former may be produced by classifying grids accord-
ing to particular approaches, e.g. by correlation methods (Schmutz
and Wanner 1998), EOF-clustering techniques (Jacobeit et al. 1998;
Beck 2000; Luterbacher et al. 2001) or assignments of grids to
predefined ‘Grosswettertypes’ (Beck 2000). Thus, each individual
pressure field strictly arrives at just one of the pressure field classes
with circulation variability being reflected in variations of fre-
quency distributions among these classes. A major drawback of
these approaches is the fact that strict boundaries between classes
might separate individual fields with substantial correspondences,
resulting in parts of the frequency changes only due to random
pressure variations.
Derivation of dynamical modes of the atmospheric circulation
is often based on various EOF techniques (e.g. Barnston and
Livezey 1987), sometimes extended into a determination of low-
dimensional regimes (Molteni et al. 1990; Corti et al. 1999) or
coupled with climatic field variables in terms of canonical modes of
variability (e.g. Beck 2000). This kind of analysis does not classify
exclusively but rather gives multiple sets of continuous time coef-
ficients for the identified modes which reproduce the original fields
according to their varying weights.
In the present context T-mode principal component analysis
(PCA) is the appropriate method since 341 monthly mean SLP
grids for January and July, respectively, for 1659–1999 have to be
analysed across 2911 grid points. The T-mode ensures a reasonable
ratio between variables and cases by taking the SLP grids as the
former, the grid points as the latter (Richman 1986). PCA implies
that resulting eigenvectors are weighted by the square root of their
eigenvalues. Thus, the weighted eigenvector components represent
the correlation coefficients between original variables and the re-
sulting principal components. In the T-mode they provide the time
coefficients of the SLP modes whose spatial patterns are determined
by the PC scores. Results are improved by orthogonal rotation
(Richman 1986), and only those rotated PCs are retained having
the greatest time coefficient among all the PCs for at least one of the
original variables (i.e. there is at least one monthly SLP grid
dominated by this mode).
This method has likewise been applied (Jacobeit et al. 2001b) to
the monthly SLP grids reconstructed back to 1780 by Jones et al.
(1999). In the former paper circulation variability has been de-
scribed in terms of cumulative anomalies of PC time coefficients
and of SLP composites referring to maximal differences within
particular SLP patterns. In contrast to that, circulation variability
for the extended period back to the mid-seventeenth century will be
evaluated by means of newly defined running indices describing
variations of PCA-derived SLP modes concerning their frequencies
as well as dynamic and climatic properties of these modes:
a. The term ‘frequency’ is not strictly applicable to PCA-derived
modes, their time coefficients rather indicate their relative impor-
tance within the original pressure fields. In order to disregard those
cases with lower importance of a particular mode, we concentrate
on those cases where its time coefficient is the leading one among all
other coefficients for the same time. We might count these cases for
a given time interval to produce a frequency term for the dominant
incidence of this mode. However, this would also disregard the
explicit information on varying importance expressed by the PC
time coefficients. The most important part of this information is
maintained by summing up all the leading time coefficients for a
particular mode for a given time interval, thus constructing a
weighted frequency term which represents the frequency of domi-
nant incidence of this mode with each case weighted by the cor-
responding time coefficient. This sum of leading time coefficients
will be determined for running 31-year periods (i.e. 1659–1689,
1660–1690,…, 1969–1999) in order to reveal mode-incidence vari-
ability efficiently (shorter periods would increase zero-frequencies,
longer periods would enhance smoothing).
b. Circulation modes not only change their incidence, but also
their internal characteristics. An important one is the flow intensity
depending on pressure gradients which may differ between different
periods of incidence of the same circulation mode. Since flow
configuration changes from one mode to another, pressure gradi-
ents have to be determined between changing locations. For an
index of intensity we calculate the pressure difference between
the moving cores of the main centres of action depending on the
dominant mode, respectively (e.g. for the zonal mode between the
pressure maximum around the Azores and the minimum near
Iceland, for the mode with an extended Russian High between this
centre and its cyclonic counterpart above the Atlantic). To describe
each mode’s flow intensity for running 31-year periods as de-
scribed, a weighted mean of pressure gradients is calculated from
all months with the same dominant mode, respectively, inserting
the leading time coefficients as weighting factors (i.e. the intensity
from an individual mode incidence enters the index according to its
relative importance).
c. Not only pressure gradients between centres of action, but
also positions of these centres may vary within the same circulation
mode thus modulating its flow pattern. This might significantly
affect particular regions: for example, the mode of winter-time
characterised by an Atlantic low-pressure system and an anticy-
clone over Russia (see later) will be linked with cold air advection
from easterly directions towards Central or even Western Europe
for westerly positions of these centres, whereas more easterly po-
sitions will imply warm air advection from westerly directions for
the same regions in between. Even without a change in flow di-
rection particular regions may be affected by spatial shifts of flow
domains as in the case of mid-latitudinal Europe with increased
cyclonic (anticyclonic) influence due to southward (northward)
shifted westerlies within the zonal circulation mode (Jacobeit et al.
2001b).
To catch within-mode variations in flow patterns we refer to the
relative vorticity given by
f ¼ dv=dx ? du=dy
with u and v as horizontal wind components along the zonal (x)
and meridional (t) axes, respectively. Instead of approximating the
partial derivatives (d) of these wind components, a simple vorticity
index has been defined based on the correlation coefficients between
each monthly SLP grid within a spatially fixed subsection of the
study area and an idealised cyclonic pressure distribution for this
subsection (Beck 2000). This index varies between +1 and –1 ac-
cording to the degree of cyclonicity or anticyclonicity within the
actual SLP grids. The subsection to which this index refers has been
fixed to the central area from 40?N to 60?N and from 10?W to 30?E
thus reflecting most of the variations in flow patterns. To describe
these variations for each circulation mode, weighted means of this
vorticity index for running 31-year periods are calculated from all
months with the same dominant mode, respectively, inserting the
leading time coefficients as weighting factors (according to the
preceding indices for frequency and intensity). Finally, the resulting
time series for each mode are standardised to have a common
scaling for the different variations of each mode concerning mean
value and amplitude. Composite maps for each mode, based on
months with standardised vorticity indices beyond +1 or –1, re-
spectively, will give an idea from the recorded variations within
each mode occurring between these contrasting subtype patterns.
d. Variations within the same circulation mode not only do
occur with respect to flow parameters (e.g. pressure gradient or
relative vorticity), but also in relation to climatic characteristics:
temperature and precipitation values experienced by a particular
region during times with dominance of the same circulation mode
will cover considerable ranges as might be indicated by widespread
instationarities in the relationships between circulation and climatic
indices (e.g. Osborn et al. 1999; Jacobeit et al. 2001a; Slonosky et al.
2001a; Pozo-Va ´ zquez et al. 2001). Thus, the gridded temperature
and precipitation data have been used to describe this variability by
calculating areally weighted averages for a given region (in this case
the continental grid boxes of Central Europe from 47?N to 54?N
and from 6?E to 14?E). From these mean values indices for each
circulation mode and for each 31-year period have been averaged
from all months with the same dominant mode, respectively,
weighting each monthly value by the leading time coefficient of the
corresponding circulation mode. Resulting variabilities are due to
different influences: circulation characteristics varying within the
Jacobeit et al.: Atmospheric circulation variability in the North-Atlantic-European area343

Page 4

same mode on the one hand (e.g. vorticity modulating precipita-
tion), varying climatic boundary conditions on the other hand (e.g.
changing SSTs modulating properties of advected air masses). To
make a first guess concerning these different influences, the running
indices for intensity and vorticity have been commonly regressed
against the temperature and precipitation indices, respectively,
giving shared variances (r2) between circulation and climate char-
acteristics for particular circulation modes (to achieve a reasonable
smoothing, these running regressions will be based on moving 60-
year periods). Unexplained variances, on the other hand, represent
climatic influences independent from the dynamic parameters
considered in this context.
4 Results
Analyses cover the period 1659–1999 thus reaching back
to the Maunder Minimum, the coldest period within the
Little Ice Age. Results will be discussed for the central
month of winter (January) and summer (July), respec-
tively.
4.1 January
4.1.1 Circulation patterns
T-mode PCA yields five circulation modes for January
if those PCs are extracted which become the dominant
mode at least for one January month during the 1659–
1999 period. Two of them, however, are the dominant
mode only for very few anomalous cases and will not
be considered hereafter. The remaining three PCs ac-
count for some 90% of the original variance and thus
represent the major circulation modes at sea-level for
January. They will be designated by abbreviations
pointing to distinct pattern characteristics (W: westerly
flow pattern; L: central low pressure centre; RH:
extended Russian High). Figure 1 gives typical SLP
distributions for these modes: instead of ordinary PC
patterns contrasting composites are shown for subs-
amples of SLP grids with the standardised vorticity
index (see section 3) being beyond +1 or –1, respec-
tively (comprising some 30% of the original grids).
These composites indicate the range in pattern config-
uration within the same circulation mode. The range
for the RH mode, however, is not recorded adequately
by this vorticity index due to an elevated degree of
internal variability. Therefore, mode RH has been di-
vided into two different patterns (easterly; westerly)
defined by the contrasting sign of a zonal index (cor-
relation coefficient of the actual SLP grid with an
idealised pressure distribution for zonal westerlies, see
Beck 2000) over the same central area for which the
vorticity index has been determined (see Sect. 3). Thus,
both patterns are associated with the same circulation
mode, but are dealt with separately in the following
analyses according to the different flow in the central
part of the study area.
The first mode (W) accounting for 37.9% of the
January SLP variance, represents the zonal circulation
regime varying internally between the reproduced
subtypes of Fig. 1a: the first one (left panel) with the
subtropical high being restricted to the Azores region
thus giving way for the westerlies to enter the whole mid-
latitudinal continent. The other subtype (right panel of
Fig. 1a) includes high pressure extending into Central
Europe thus diverting the westerlies towards higher
latitudes over the British Isles and Scandinavia. The
more our mode-dependent index of vorticity changes to
positive (negative) values (see later), the more the actual
flow configuration within mode W approaches the left
(right) subtype pattern.
Mode L accounting for 25.6% of the original vari-
ance, depicts a strong Atlantic Low as the only major
centre of action, deepening and retreating farther west in
the second subtype (right panel) of Fig. 1b.
Mode RH, distinguished into easterly and westerly
patterns with 8.5% and 17.6% of explained variance,
respectively, represents a circulation regime with varying
influence of a distinct Russian high. Among the easterly
patterns, known for cold anomalies in central Europe,
flow configurations vary between the subtypes of
Fig. 1d: the first one (left panel) with the Russian High
at more northern latitudes and a marked central Medi-
terranean Low, another one (right panel) with the high-
pressure influence extending towards Central and
Western Europe. The westerly patterns (Fig. 1c), asso-
ciated with increased temperatures in western parts of
Europe, show an enhanced influence of the Atlantic Low
pressure system modulated by the differently receding
impact of the Russian High reflected in the two subtypes
of Fig. 1c.
4.1.2 Frequency variations
The running frequency indices as explained in Sect. 3 are
given in Fig. 2. The upper panel shows variations for
the PCA-derived modes (i.e. without splitting mode RH
into westerly and easterly patterns) indicating general
antiphase variations between the zonal westerlies and
the other two modes. Correlation coefficients (W/RH
and W/L) do not exceed moderate levels (–0.61 and
–0.69, respectively), but on long-term time scales periods
of increased versus decreased frequencies approximate
with phase shiftings of several decades, however,
between modes RH and L. Since the cyclical variations
operate on multi-decadal time scales, the whole avail-
able period of 340 years is still too short for evaluating
persistencies of these low-frequency cycles. Mode W, for
example, just has three periods of enhanced importance:
during the first half of the eighteenth century and
around the beginnings of the nineteenth and twentieth
centuries.
Remarkably, there were two periods when mode RH
distinctly exceeded mode W (around the 1860s and the
1930s). The former period, however, coincides with
positive anomalies in the NAO which started to accu-
mulate during winter since the 1850s (Jacobeit et al.
344Jacobeit et al.: Atmospheric circulation variability in the North-Atlantic-European area

Page 5

2001a). This seeming contradiction is resolved by the
lower panel of Fig. 2 indicating that these RH maxima
are due to the westerly pattern of this mode which
generally dominates against the easterly pattern since the
mid-nineteenth century. During earlier times, however,
RHeastalso dominated for particular periods, especially
during cold episodes of the Little Ice Age such as the
Late Maunder Minimum around the turn of the eigh-
teenth century (Luterbacher et al. 2001) or during the
first half of the nineteenth century (Wanner et al. 2000).
Fig. 1a–d
(hPa) of January SLP modes
1659–1999. Left (right) panel:
weighted means of those SLP
grids with Central European
standardised vorticity deviation
index for each SLP mode being
greater than +1 (lower than –1)
Subtype composites
Jacobeit et al.: Atmospheric circulation variability in the North-Atlantic-European area 345

Page 6

Thus, an important change in circulation is indicated for
the mid-nineteenth century around the end of the Little
Ice Age period.
4.1.3 Within-mode variations
The running indices describing variations in vorticity,
intensity, temperature and precipitation for the particu-
lar SLP modes (see Sect. 3) are shown in Fig. 3. For
mode W (Fig. 3a) all indices reached higher values dur-
ing the twentieth century than most of the time before,
i.e. the zonal circulation mode formerly occurred with
less cyclonic vorticity, lower intensity, temperature and
precipitation than during much of the last century. The
main transitions, however, took place at different times:
already starting around 1800 for intensity and tempera-
ture (r = 0.68) with the former reaching modern levels by
the second half of the nineteenth century, the latter
rapidly peaking around 1930. At the same time vorticity
and precipitation (r = 0.76) reached their first maxima
after strongly increasing since the end of the nineteenth
century. As a whole, long-term changes within mode W
extended from a pronounced period around 1800 with
below-average indices until the 1930s with above-average
indices. The most recent development of the last few
decades reveals decreasing vorticity implying a tendency
towards northward shifted westerlies with increased
Fig. 3a–d
tions for January SLP modes
since 1659 (weighted means for
moving 31-year periods). Vort:
standardised deviations of
Central European relative vor-
ticity. Int: pressure gradient
(hPa/1000 km) between the
mode-dependent centres of
action. T: Central European
temperature (?C) during occur-
rence of the specified SLP
mode. R: Central European
precipitation (mm) during
occurrence of the specified
SLP mode. Central lines
represent the long-term
averages (1659–1999), respec-
tively. Standard deviations of
all parameters are equally
scaled on the vertical axes
Within-mode varia-
Fig. 2
expressed as sums of leading time coefficients for moving 31-year
periods (values are located at the centre of the corresponding
period). The lower panel decomposes the time series for mode RH
into subseries for its westerly and easterly patterns
The relative importance of January SLP modes since 1659
346Jacobeit et al.: Atmospheric circulation variability in the North-Atlantic-European area

Page 7

anticyclonicity to the south as already reported for the
Alpine region by Wanner et al. (2000). However, this
recent shift maintains increasing values for intensity and
temperature in contrast to earlier times.
Within-mode changes do not merely reflect general
changes in climatic boundary conditions but may well
differ between different circulation modes. This becomes
evident when looking at the westerly pattern of mode
RH (Fig. 3c) where only the intensity index reveals a
significant tendency to increase towards the mid-twen-
tieth century. The weak trends towards higher temper-
atures (across the whole period) and more anticyclonic
subtypes (since ?1800) are not statistically significant
and superimposed by decadal-scale variations of differ-
ent amplitudes and wavelengths. However, comparing
the indices of Fig. 3c for the last two periods of maxi-
mum frequency (see Fig. 2) clearly indicates that all of
them had higher values during the first decades of the
twentieth century than during the second half of the
nineteenth century. However, this increase was not part
of systematic long-term changes as revealed for mode W.
More prominent are various anomaly periods with the
most important one around 1800 when the maximum of
cyclonic subtypes was reached, linked with elevated
precipitation amounts and, unlike the linkage for mode
W, with below-average temperatures in Central Europe.
For the easterly pattern of mode RH (Fig. 3d) such
anomaly periods (discernible by maxima or minima in
the vorticity index) coincide with reduced frequencies,
thus being based on just a few particular cases, which
will not be discussed further. However, there are some
distinct tendencies in the climatic parameters: a sus-
tained warming of mode RHeastacross the nineteenth
century as well as a wetness increase which started some
50 years earlier, but changed to a downward trend
during the 1870s. This evolution clearly contrasts with
mode W (Fig. 3a) thus indicating that the general in-
crease in Central European winter precipitation during
the last century (Beck et al. 2001) does not apply to each
mode’s typical precipitation amounts. A common fea-
ture of modes W and RHeast, however, are the higher
pressure gradients of recent times compared with the
whole of the earlier period.
This does not apply to mode L whose modern
intensity still varies around medium values (Fig. 3b).
However, long-term periods may be discerned with re-
spect to vorticity subtypes developing a pronounced
cyclonic maximum during the second half of the eigh-
teenth century, falling afterwards to enter a sustained re-
increase during the nineteenth century and proceeding to
the modern evolution towards less cyclonic conditions.
These long-term variations are inversely reflected by the
thermal evolution of mode L resulting in some clear-cut
anomaly periods with pronounced cyclonicity, enhanced
precipitation and, related to mean values of mode L,
below-average temperatures in Central Europe: most
conspicuously after 1750, somewhat less striking around
1900. Thus, warm advection from the southwest is
rather linked with more anticyclonic subtypes whereas
enhanced cyclonic conditions imply the overturning of
different air masses.
4.2 July
4.2.1 Circulation patterns
One of the five PCs extracted becomes the dominant
mode for only very few July months. Thus, the July
circulation may be described by four principal modes of
variation explaining some 97% of the original variance.
Once more, abbreviations should point to distinct pat-
tern characteristics (W: westerly flow pattern; CY: major
cyclonic centre; AR: Atlantic ridge pattern; CR: conti-
nental ridge pattern). Figure 4 shows typical pressure
distributions for these modes in terms of composites for
subsamples of SLP grids with the standardised vorticity
index being beyond +1 or –1, respectively (comprising
some 30% of the original grids).
Mode W accounting for 28.1% of the original vari-
ance represents a summer zonal circulation varying be-
tween opposite subtypes (Fig. 4a): a stronger subpolar
low with restrained anticyclonic influence to its south
(left panel), and a weakened low near Iceland with a
zonal ridge of high pressure extending across central
Europe (right panel). Mode CY (26.4% of explained
variance) shows similar behaviour with the cyclonic
centre of action, however, being located farther eastward
over Northern Europe for both subtypes (Fig. 4b).
The following modes of July extend anticyclonic cir-
culations towards higher latitudes: mode AR (23.9% of
explained variance) with a strong ridge from the Azores
towards the northeast including subtypes of different
extension between the North Sea and the Baltic region
(Fig. 4c); mode CR (18.9% of explained variance) with
high pressure over Fennoscandia as part of an anticy-
clonic domain over the continent including subtypes
with different regional prevalence between Northern and
Central Europe (Fig. 4d).
4.2.2 Frequency variations
Most outstanding in Fig. 5 is the unique maximum of
mode W at the end of the eighteenth century. In con-
trast to common suggestions this did not coincide with
Central European summer temperatures below, but
rather distinctly above the long-term average during
this period (Jacobeit et al. 1998; Beck 2000) thus
pointing to important changes within this mode dis-
cussed later.
Mode AR, in contrast, was at its minimum around
the 1790s and underwent a long-term increase to its re-
cent maximum whereas mode CY dropped to its mini-
mum during the last few decades. This means a general
shift towards anticyclonic steering over the western part
of the European continent with its major develop-
ment occurring since the mid-twentieth century as an
Jacobeit et al.: Atmospheric circulation variability in the North-Atlantic-European area 347

Page 8

extraordinary phenomenon during the last 340 years
(Fig. 5). Periods with more pronounced cyclonic char-
acter repeatedly occurred during the historical past:
around the end of the nineteenth century, around the
1830s and especially during the first half of the eigh-
teenth century when a distinct maximum of mode
CY coincided with the falling-off of mode CR (Fig. 5).
Its preceding maximum, however, only has restricted
Fig. 4a–d
(hPa) of July SLP modes 1659–
1999. Left (right) panel:
weighted means of those SLP
grids with Central European
standardized vorticity deviation
index for each SLP mode being
greater than +1 (lower than –1)
Subtype composites
348 Jacobeit et al.: Atmospheric circulation variability in the North-Atlantic-European area

Page 9

validity since the reconstruction skill of gridded SLP
during summer is still rather poor at least until the end
of the seventeenth century (Luterbacher et al. 2002a).
4.2.3 Within-mode variations
For mode W (Fig. 6a) distinct antiphase correlations are
evident between the vorticity parameter and both inten-
sity (r = –0.68) and Central European pattern tempera-
ture (r = –0.74). Relationships with pattern precipitation
are more variable due to independent variations in spe-
cific humidity and convective activity. Most prominent
are several anomaly periods: on the one hand character-
ised by cyclonic subtypes with below-average tempera-
turessuchasduringthefirsthalfoftheeighteenthcentury
or, less pronounced, around 1850 and 1900. On the other
hand, the reverse case has its most distinct occurrence at
the end of the eighteenth century. This was just the time
with increased frequency of mode W (see Fig. 5) coin-
ciding with the period mentioned of marked summer
warmth in Central Europe. This warmth was obviously
not due to increased frequencies of modes AR or CR, but
rather linked to anticyclonic patterns of mode W (Fig. 4)
with strengthened pressure gradients to the north. These
conditions contributed to the elevated temperatures
indicated for mode W around this period (Fig. 6a).
Most outstanding for mode CY is the strong cyclonic
anomaly linked with increased precipitation at the
Fig. 5
expressed as sums of leading time coefficients for moving 31-year
periods (values are located at the centre of the corresponding
period)
The relative importance of July SLP modes since 1659
Fig. 6a–d
variations for July SLP modes
since 1659 (weighted means for
moving 31-year periods).
Abbreviations as in Fig. 3
Within-mode
Jacobeit et al.: Atmospheric circulation variability in the North-Atlantic-European area349

Page 10

beginning of the second half of the eighteenth century
(Fig. 6b). During this time, however, mode CY passed
through a frequency minimum (Fig. 5) whereas the
preceding maximum was rather linked with anticyclonic
subtype preferences (Fig. 6b). Remarkably, the recent
decline in frequency of mode CY since the 1950s is ac-
companied by unique declines of its intensity and tem-
perature until the closing decades of the twentieth
century.
The striking intensity anomaly of mode AR (Fig. 6c)
at the end of the eighteenth century becomes less im-
portant in view of the simultaneous frequency minimum
(Fig. 5) whereas the pronounced anticyclonic conditions
with elevated temperatures and reduced precipitation at
the beginning of the eighteenth century occurred with
increased frequency. From this period onward until the
first half of the twentieth century a long-term trend to-
wards cyclonic subtypes was superimposed upon the
decadal variability in the vorticity parameter (Fig. 6c)
changing to the opposite trend (decreasing vorticity)
during the period of increasing AR frequency since the
1950s (Fig. 5). The internal warming of mode AR after
the eighteenth century’s cooling was only interrupted
around 1900 roughly concomitant with a marked vor-
ticity increase (Fig. 6c).
For mode CR most of the index maxima and mini-
ma of Fig. 6d coincide with rather low frequencies
(Fig. 5), the twentieth century still showing the most
coherent tendencies: towards anticyclonic subtypes with
distinctly declining precipitation, towards higher inten-
sity which only drops again with the rather late internal
warming.
5 Discussion and conclusions
Atmospheric circulation modes derived for the last 340
years by multivariate analyses of reconstructed SLP
grids generally correspond to modes already determined
by similar analyses extending back to the end of the
eighteenth century (Slonosky et al. 2000; Jacobeit et al.
2001b; Beck et al. 2001). A major advance in this study
on past circulation variability, however, has been
achieved by considering not only frequency variations
(Figs. 2 and 5), but also internal changes within these
modes. In contrast to Jacobeit et al. (2001b) who showed
such changes only for selected periods during the past
two centuries, within-mode variability is documented
continuously throughout the past 340 years in Figs. 3
and 6. Furthermore, these figures not only refer to
mode-dependent temperature and precipitation charac-
teristics as outlined by Beck et al. (2001) for SLP pat-
terns since 1780, but also include indices of within-mode
variability in pressure gradients and relative vorticity.
Thus, variations between contrasting subtypes for each
mode (Figs. 1 and 3) could be continuously identified
since the mid-seventeenth century.
Internal variationswithin
modes are major dynamical factors in the context of
particularcirculation
historical climatic changes. This may be seen from the
results for the zonal circulation mode in January. Since
frequency changes rather occurred as fluctuations on
multi-decadal time scales (Fig. 2), the climate transition
from the Little Ice Age to the following period pro-
ceeded mainly by way of long-term within-mode changes
(Fig. 3a). Some of these had already started by 1800,
increasing pressure gradients and rising mode-W-tem-
perature, whereas mode-W-precipitation and relative
vorticity dropped once more during the second half of
the nineteenth century before increasing distinctly af-
terwards. The whole within-mode transition concluded
around 1930 when all parameters had reached well
above-average values.
Further evidence refers to the January mode RH
characterised by a Russian High of variable extension
to the west. Based on the time coefficient of a corre-
sponding EOF derived from SLP data since 1774,
Slonosky et al. (2000) refer to a period of more intense
meridional circulation from 1822 to 1870. This seems
also to be indicated by RH frequencies exceeding those
of mode W around the mid-nineteenth century (Fig. 2,
upper panel). Further information, however, depicts
that just within this ‘meridional period’ important
changes took place around 1850: from dominating
easterly to dominating westerly patterns of mode RH
(Fig. 2, lower panel), concomitant with a major turning
point in the long-term evolution of cumulative NAO
anomalies from negative to positive predominances
(Jacobeit et al. 2001a). This turning point is also
reflected in the transition of mode W intensities from
below- to above-average values around the mid-nine-
teenth century (Fig. 3a). At the same time the reorga-
nisation of the atmospheric circulation towards the end
of the Little Ice Age implied prevailingly eastward
shifted positions within mode RH leading to a pre-
dominance transition from easterly to westerly pat-
terns.
The main long-term evolution during July may be
seen as a general increase of anticyclonic conditions
during the last two centuries indicated by opposite fre-
quency changes of modes AR and CY (Fig. 5) and by
general shifts towards anticyclonic subtypes (Fig. 6) for
modes W (since the mid-nineteenth century), CR and
AR (during the last 80 and 50 years, respectively). Mode
CY maintains the highest rainfall index, but declines in
frequency. The generally rising anticyclonicity is also
reflected within the European trend atlas 1891–1990
(Scho ¨ nwiese et al. 1993) showing increasing pressure for
Central Europe during July. Opposite conditions pre-
vailed during the first half of the eighteenth century (a
strong frequency maximum of mode CY and preferred
cyclonic subtypes of mode W) before a distinct anomaly
period occurred towards the end of the eighteenth cen-
tury with a striking frequency maximum of mode W
preferring anticyclonic subtypes at the same time
(Figs. 5 and 6a). This within-mode variation, however, is
not yet able to explain major parts of the elevated mode-
W-temperatures around this time which is known from
350Jacobeit et al.: Atmospheric circulation variability in the North-Atlantic-European area

Page 11

independent data as a period of marked summer warmth
in Central Europe (Beck 2000). Figure 7a clearly shows
that variances of mode-W-temperature explained by
variations of circulation parameters just broke down
during this period. This means that climatic boundary
conditions independent from variations within mode W
additionally favoured higher summer temperatures
around the turn of the eighteenth century.
Thus, in general, mode-dependent variations of cli-
matic parameters have to be distinguished into one part
explained by varying circulation characteristics, and
another part due to varying climatic boundary condi-
tions. Both parts change with time (Fig. 7a) and differ
between modes. Another example referring to precipi-
tation changes within mode CY (Fig. 7b) reveals that
they are mostly well explained by changes of vorticity
and intensity, including the strong cyclonic anomaly at
the beginning of the second half of the eighteenth cen-
tury.
A final look at present conditions in view of the
variations since the mid-seventeenth century reveals that
only a few of the recent circulation phenomena may be
really unusual or unprecedented. Even the strong in-
crease in the winter-time westerly circulation during the
last decades does not appear extraordinary considering
the low-frequency cycle indicated for mode W in Janu-
ary (Fig. 2): if this natural variation on the time scale
of 80–100 years should continue, a further increase
of mode W during the forthcoming decades may be
expected independently from man-made impacts due to
the enhanced greenhouse forcing (note that final values
of Fig. 2 represent the period from the late 1960s to the
late 1990s when zonal circulation just proceeded
from below- to well above-average values). Unusual
conditions, however, are indicated recently for some
mode-W-parameters (Fig. 3a): despite the tendency of
decreasing vorticity during the last few decades, tem-
perature and intensity continue to increase indicating a
warming also of subtypes with rising anticyclonicity
above southern Central Europe. This evolution reverses
the positive relation between vorticity and temperature
for January mode W and defines novel conditions within
the last 340 years.
During July, the long-term evolution of increasing
anticyclonicity has strengthened during the last 50 years
becoming recently a unique phenomenon within the last
centuries, as may be seen from the diverging frequency
changes of modes CY and AR (Fig. 5) and from the
distinctly declining precipitation values of mode CR
(Fig. 6d). Together with recently preferred anticyclonic
subtypes of mode W (Fig. 6a) this caused an unprece-
dented decline in Central European July precipitation
during the second half of the twentieth century (Beck
et al. 2001).
Circulation analyses based on reconstructed SLP
grids for several centuries may be extended to further
investigations, too, e.g. to dynamic studies on regional
climate variability as far as SLP data independent from
climatic predictors are available (Beck et al. 2001). An-
other approach recently applied by Sturm et al. (2001)
and Wanner et al. (submitted, 2002) refers to relation-
ships between historical flood events and atmospheric
circulation modes. This kind of analysis directs dynamic
studies even towards the social impacts of extreme
events.
Acknowledgements This work has been supported by the German
Science Foundation (DFG) under grant JA 831 and by the Swiss
National Science Foundation (SNSF) under grant number
11-52786-97 as well as by the National Centre of Competence in
Research (NCCR) in Climate funded by SNSF.
References
Appenzeller C, Stocker TF, Anklin M (1998) North Atlantic Os-
cillation dynamics recorded in Greenland ice cores. Science 282:
446–449
Ba ¨ rring L, Jo ¨ nsson P, Achberger C, Ekstro ¨ m M, Alexandersson H
(1999) The Lund instrumental record of meteorological obser-
vations: reconstruction of monthly sea-level pressure 1780–
1997. Int J Climatol 19: 1427–1443
Barnston AG, Livezey RE (1987) Classification, seasonality and
persistence of low frequency atmospheric circulation patterns.
Mon Weather Rev 115: 1083–1126
Beck C (2000) Zirkulationsdynamische Variabilita ¨ t im Bereich
Nordatlantik-Europa seit 1780. Wu ¨ rzburger Geographische
Arbeiten 95, pp 350
Beck C, Jacobeit J, Philipp A (2001) Variability of North-Atlantic-
European circulation patterns since 1780 and corresponding
variations in Central European climate. In: Brunet-India M,
Lo ´ pez-Bonillo D (eds) Detecting and modelling regional cli-
mate change. Springer, Berlin Heidelberg New York, pp 321–
332
Cook ER, D’Arrigo RD, Briffa KR (1998) The North Atlantic
Oscillation and its expression in circum-Atlantic tree-ring
chronologies from North America and Europe. The Holocene
8: 9–17
Cook ER, D’Arrigo RD, Mann M (2002) A well-verified, multi-
proxy reconstruction of the winter North Atlantic Oscillation
Index since AD 1400. J Clim 15: 1754–1764
Fig. 7a, b
respectively, explained for moving 60-year periods by vorticity and
intensity parameters (multiple r2, grey areas) of a July SLP mode
W; b July SLP mode CY. T: Central European temperature (?C) of
July SLP mode W. R: Central European precipitation (mm) of July
SLP mode CY
Variances of T (a, bold line) and R (b, bold line),
Jacobeit et al.: Atmospheric circulation variability in the North-Atlantic-European area 351

Page 12

Cullen HM, D’Arrigo RD, Cook ER, Mann M (2000) Multiproxy
reconstructions of the North Atlantic Oscillation. Paleocea-
nography 16: 27–39
Corti S, Molteni F, Palmer TN (1999) Signature of recent climate
change in frequencies of natural atmospheric circulation re-
gimes. Nature 398: 799–802
Garcia R, Gimeno L, Hernandez E, Prieto R, Ribera P (2000)
Reconstructing the North Atlantic atmospheric circulation in
the 16th, 17th and 18th centuries from historical sources. Clim
Res 14: 147–151
Glaser R (2001) Klimageschichte Mitteleuropas. 1000 Jahre Wet-
ter, Klima, Katastrophen. Wiss. Buchgesellschaft, Darmstadt,
pp 227
Glueck MF, Stockton CW (2001) Reconstruction of the North
Atlantic Oscillation, 1429–1983. Int J Climatol 21: 1453–1465
Hurrell JW (1995) Decadal trends in the North Atlantic Oscilla-
tion: regional temperatures and precipitation. Science 269: 676–
679
Jacobeit J, Beck C, Philipp A (1998) Annual to decadal variability
in climate in Europe – objectives and results of the German
contribution to the European climate research project AD-
VICE. Wu ¨ rzburger Geographische Manuskripte 43, pp 163
Jacobeit J, Wanner H, Koslowski G, Gudd M (1999) European
surface pressure patterns for months with outstanding climatic
anomalies during the sixteenth century. Clim Change 43: 201–
221
Jacobeit J, Jo ¨ nsson P, Ba ¨ rring L, Beck C, Ekstro ¨ m M (2001a)
Zonal indices for Europe 1780–1995 and running correlations
with temperature. Clim Change 48: 219–241
Jacobeit J, Jones PD, Davies TD, Beck C (2001b) Circulation
changes in Europe since the 1780s. In: Jones PD, Ogilvie A,
Davies TD, Briffa KR (eds) History and climate: memories of
the future? Kluwer, New York, pp 79–99
Jones PD, Wigley TM, Briffa KR (1987) Monthly mean pressure
reconstruction for Europe (back to 1780) and North-America
(back to 1858). DOE Technical Report TR37, US Department
of Energy, Washington D.C., pp 99
Jones PD, Jonsson T, Wheeler D (1997) Extension of the North
Atlantic Oscillation using early instrumental pressure observa-
tions from Gibraltar and southwest Iceland. Int J Climatol 17:
1433–1450
Jones PD, et al (1999) Monthly mean pressure reconstruction for
Europe 1780–1995. Int J Climatol 19: 347–364
Kalnay E, et al (1996) The NCEP/NCAR 40-Year Reanalysis
Project. Bull Am Meteorol Soc 77: 437–471
Kington JA (1988) The weather of the 1780s over Europe. Cam-
bridge University press, Cambridge, UK, pp 166
Kistler R, et al (2001) The NCEP-NCAR 50-year Reanalysis:
monthly means CD-ROM and documentation. Bull Am
Meteorol Soc 82: 247–267
Lamb HH, Johnson AI (1966) Secular variations of the atmo-
spheric circulation since 1750. Geophys Mem 110, pp 125
Luterbacher J, Schmutz C, Gyalistras D, Xoplaki E, Wanner H
(1999) Reconstruction of monthly NAO and EU indices back to
AD 1675. Geophys Res Lett 26: 2745–2748
Luterbacher J, Rickli R, Xoplaki E, Tinguely C, Beck C, Pfister C,
Wanner H (2001) The Late Maunder Minimum (1675–1715) –
a key period for studying decadal scale climatic change in
Europe. Clim Change 49: 441–462
Luterbacher J, et al (2002a) Reconstruction of sea level pressure
fields over the eastern North Atlantic and Europe back to 1500.
Clim Dyn 18: 545–561
Luterbacher J, Xoplaki E, Dietrich D, Jones PD, Davies TD, Portis
D, Gonzalez-Rouco JF, von Storch H, Gyalistras D, Casty C,
Wanner H (2002b) Extending North Atlantic Oscillation re-
constructions back to AD 1500. Atmos Sci Lett 2: 114–124
Molteni F, Tibaldi S, Palmer TN (1990) Regimes in the wintertime
circulation over northern extratropics. I: observational evi-
dence. Q J R Meteorol Soc 116: 31–67
New M, Hulme M, Jones PD (2000) Representing twentieth cen-
tury space-time climate variability. II: development of 1901–
1998 monthly grids of terrestrial surface climate. J Clim 13:
2217–2238
Osborn TJ, Briffa KR, Tett SFB, Jones PD, Trigo RM (1999)
Evaluation of the North Atlantic Oscillation as simulated by a
coupled climate model. Clim Dyn 15: 685–702
Pfister C (1999) Wetternachhersage. 500 Jahre Klimavariationen
und Naturkatastrophen 1496–1995. Bern Stuttgart Wien, pp
304
Portis DH, Walsh JE, El Hamly M, Lamb PJ (2001) Seasonality of
the North Atlantic Oscillation. J Clim 14: 2069–2078
Pozo-Va ´ zquez D, Esteban-Parra MJ, Rodrigo FS, Castro-Diez Y
(2000) An analysis of the variability of the North Atlantic Os-
cillation in the time and the frequency domains. Int J Climatol
20: 1675–1692
Pozo-Va ´ zquez D, Esteban-Parra MJ, Rodrigo FS, Castro-Diez Y
(2001) A study of NAO variability and its possible non-linear
influences on European surface temperature. Clim Dyn 17: 701–
715
Richman MB (1986) Rotation of principal components. J Climatol
6: 293–335
Rodrigo FS, et al (2001) A reconstruction of the winter North
Atlantic Oscillation Index back to 1501 using documentary data
in Southern Spain. J Geophys Res 106: 14,805–14,818
Schmutz C, Wanner H (1998) Low frequency variability of atmo-
spheric circulation over Europe between 1785 and 1994.
Erdkunde 52: 81–94
Schmutz C, Luterbacher J, Gyalistras D, Xoplaki E, Wanner H
(2000) Can we trust proxy-based NAO index reconstructions?
Geophys Res Lett 27: 1135–1138
Scho ¨ nwiese CD, Rapp J, Fuchs T, Denhard M (1993) Klimatrend-
Atlas Europa 1891–1990. Berichte des Zentrums fu ¨ r Umweltf-
orschung der J.W.Goethe-University Frankfurt am Main,
Germany, 20
Slonosky VC, Jones PD, Davies TD (2000) Variability of the sur-
face atmospheric circulation over Europe, 1774–1995. Int J
Climatol 20: 1875–1897
Slonosky VC, Jones PD, Davies TD (2001a) Atmospheric circu-
lation and surface temperature in Europe from the 18th century
to 1995. Int J Climatol 21: 63–75
Slonosky VC, Jones PD, Davies TD (2001b) Instrumental pressure
observation from the 17th and 18th centuries: London and
Paris. Int J Climatol 21: 285–298
Sturm K, Glaser R, Jacobeit J, Deutsch M, Brazdil R, Pfister C,
Luterbacher J, Wanner H (2001) Hochwasser in Mitteleuropa
seit 1500 und ihre Beziehung zur atmospha ¨ rischen Zirkulation.
Petermanns Geogr Mitt 145(6): 14–23
Wanner H, et al (1994) Synoptic interpretation of monthly weather
maps for the Late Maunder Minimum (1675–1704). Paleoclim
Res 13: 401–424
Wanner H, et al (1995) Wintertime European circulation patterns
during the Late Maunder Minimum cooling period (1675–
1704). Theor Appl Climatol 51: 159–165
Wanner H, Gyalistras D, Luterbacher J, Rickli R, Salvisberg E,
Schmutz C (2000) Klimawandel im Schweizer Alpenraum. Vdf
Hochschulverlag AG, Zu ¨ rich, pp 294
352 Jacobeit et al.: Atmospheric circulation variability in the North-Atlantic-European area