138
J. Luterbacher and E. Xop\aki
were calculated and the higher order EOFs have been discarded. The subsequent
multivariate regression was performed linking the predictors to the predictands, i.e.
the determination of the linear transfer functions. Then, the quality of our
reconstructions was tested by applying each of the 320 multivariate regression
models to independent data from the verification period 1961-1990.
Finally, the EOFs for the predictors and predictands were derived over the entire
90 years from 1901-1990 and the subsequent principal components regression
analysis performed and applied to the predictor data giving the temperature and
precipitation estimates back to AD 1500. For a full mathematical treatment of the
reconstruction method, the reader is referred to Luterbacher et al. (200lc).
The quality ofthe reconstructions for selected anomalous winters was tested grid
point by grid point over the whole region (in total 5829 grid points including the
area 30
c
N-70
c
N; 30
c
W-40
C
E) by applying each of the 320 statistical models fitted
during the period 1901-1960 and verified over the verification period 1961-1990.
The statistical measure we used for the quality check of the reconstructions was the
Reduction of Error (RE) (Lorenz 1956). RE ranges from + 1 (perfect agreement
between reconstructions and analysed fields) to -co with RE = 0 no better than
climatology (i.e. the mean of the climatological data in the calibration period), RE
> 0 better than the calibration mean and RE < 0 no useful information in the
reconstructions.
Since the Mediterranean area is of major interest in this study, we derived a time
series of averaged temperature and precipitation back to AD 1500 by taking into
account the 2159 grid points in the region 30
c
N-4TN and lOW-40cE. Therefore,
we considered the larger Mediterranean area as one region despite the known
climatic differences in the western, central and eastern Mediterranean regions. Here
we only show the results for the winter season (DJF).
Simple correlation analyses has been used to study the relationship between the
winter temperature and precipitation series and the independent reconstruction of
the winter NAOI by Cook et al. (2001), who used tree-ring data from Europe and the
U.S.A. and ice core data from Greenland. Cook's et al. (2001) NAO reconstructions
are entirely independent of our reconstructions, i.e., they include no common
predictor data.
Based on this long-term temperature and precipitation time series, we selected
the coldest, warmest, driest and wettest winters over the defined region for the
reconstruction period AD 1500 to 1900. These are shown together with their
departures from the long-term 1901-1995 winter mean in order to get an idea about
the manifestation of the anomalies over the whole of Europe, North Africa and the
Near East. In addition, the reconstruction quality (RE) (see description above) for
these four extreme winters will be presented to have an indication where meaningful
reconstructions can be expected. The corresponding reconstructed SLP fields, their
anomalies and their reconstruction quality are given as well in order to interpret the
temperature and precipitation anomalies in terms of the large-scale atmospheric
circulation.
J. Luterbacher and E. Xop\aki
were calculated and the higher order EOFs have been discarded. The subsequent
multivariate regression was performed linking the predictors to the predictands, i.e.
the determination of the linear transfer functions. Then, the quality of our
reconstructions was tested by applying each of the 320 multivariate regression
models to independent data from the verification period 1961-1990.
Finally, the EOFs for the predictors and predictands were derived over the entire
90 years from 1901-1990 and the subsequent principal components regression
analysis performed and applied to the predictor data giving the temperature and
precipitation estimates back to AD 1500. For a full mathematical treatment of the
reconstruction method, the reader is referred to Luterbacher et al. (200lc).
The quality ofthe reconstructions for selected anomalous winters was tested grid
point by grid point over the whole region (in total 5829 grid points including the
area 30
c
N-70
c
N; 30
c
W-40
C
E) by applying each of the 320 statistical models fitted
during the period 1901-1960 and verified over the verification period 1961-1990.
The statistical measure we used for the quality check of the reconstructions was the
Reduction of Error (RE) (Lorenz 1956). RE ranges from + 1 (perfect agreement
between reconstructions and analysed fields) to -co with RE = 0 no better than
climatology (i.e. the mean of the climatological data in the calibration period), RE
> 0 better than the calibration mean and RE < 0 no useful information in the
reconstructions.
Since the Mediterranean area is of major interest in this study, we derived a time
series of averaged temperature and precipitation back to AD 1500 by taking into
account the 2159 grid points in the region 30
c
N-4TN and lOW-40cE. Therefore,
we considered the larger Mediterranean area as one region despite the known
climatic differences in the western, central and eastern Mediterranean regions. Here
we only show the results for the winter season (DJF).
Simple correlation analyses has been used to study the relationship between the
winter temperature and precipitation series and the independent reconstruction of
the winter NAOI by Cook et al. (2001), who used tree-ring data from Europe and the
U.S.A. and ice core data from Greenland. Cook's et al. (2001) NAO reconstructions
are entirely independent of our reconstructions, i.e., they include no common
predictor data.
Based on this long-term temperature and precipitation time series, we selected
the coldest, warmest, driest and wettest winters over the defined region for the
reconstruction period AD 1500 to 1900. These are shown together with their
departures from the long-term 1901-1995 winter mean in order to get an idea about
the manifestation of the anomalies over the whole of Europe, North Africa and the
Near East. In addition, the reconstruction quality (RE) (see description above) for
these four extreme winters will be presented to have an indication where meaningful
reconstructions can be expected. The corresponding reconstructed SLP fields, their
anomalies and their reconstruction quality are given as well in order to interpret the
temperature and precipitation anomalies in terms of the large-scale atmospheric
circulation.