SOO-year Winter Temperature and Precipitation Variability over the Mediterranean Area
147
only slightly significant during the last centuries.
4 Discussion
The averaged winter Mediterranean temperature and precipitation time series
(Figures 2 and 3) clearly showed periods with extended dry/wet and warm/cold
periods. A physical and climatological interpretation of these time series requires
information about their quality. The model performance of these reconstructions
over the whole of Europe and North Africa is discussed in detail in Luterbacher et
al. (200Ia) and Luterbacher et al. (200lc). For the Mediterranean as a whole, the
reconstruction quality increases from the 16 th to the 20 th century gradually. The
increase of model performance is mainly a function of the number and the
distribution of the predictors. For temperature generally more trustworthy estimates
were obtained.
The overall model performance expressed by the mean RE (averaged over 2159
gridpoints) gives an indication how well, on average, the temperature and
precipitation fields are reconstructed over the whole area. The time series indicate
rather low RE values for the seasonal (AD 1500 to AD 1658) reconstructions,
though still positive.
This mainly reflects the low number of predictors available for this early period.
For the Mediterranean area there is only the precipitation predictor from Andalusia
(Southern Spain, Rodrigo et al. 2001) from 1501 onwards available. Therefore, the
averaged positive RE values for the early reconstructions are an indication that
remote predictors (from Central and Eastern Europe) can significantly contribute to
skillful temperature and precipitation reconstructions over the larger Mediterranean
area.
From the beginning of the 18 th century, with more temperature and precipitation
predictors, the quality of the monthly reconstructions increases. Provisional results
of time-dependent uncertainty ranges of the estimated temperature and precipitation
series reveal similar performance than of e.g. Briffa et al. (200 I a) which are based
on tree-ring data. The current level of uncertainty in the first centuries can be
reduced with the inclusion of further new early instrumental and proxy data from
areas sensitive to the Mediterranean area.
The spatial model performance gives insight in which areas reconstructions are
trustworthy. This has been demonstrated for the four winter extremes (Figures 4 and
5). Although the overall model performance for the wettest and driest winters
implies low reconstruction quality (not shown), there are regions within the
Mediterranean with skillful reconstructions. Therefore both, the overall and the
spatial model performance have to be considered when reconstructed temperature
and precipitation fields are climatologically interpreted.
A large fraction of precipitation and temperature variability can be explained by
large-scale circulation anomalies when averaging in space and time over adequately
large enough intervals (Trenberth 1990; Xu 1993; Qian et al. 2000; Xoplaki et al.
147
only slightly significant during the last centuries.
4 Discussion
The averaged winter Mediterranean temperature and precipitation time series
(Figures 2 and 3) clearly showed periods with extended dry/wet and warm/cold
periods. A physical and climatological interpretation of these time series requires
information about their quality. The model performance of these reconstructions
over the whole of Europe and North Africa is discussed in detail in Luterbacher et
al. (200Ia) and Luterbacher et al. (200lc). For the Mediterranean as a whole, the
reconstruction quality increases from the 16 th to the 20 th century gradually. The
increase of model performance is mainly a function of the number and the
distribution of the predictors. For temperature generally more trustworthy estimates
were obtained.
The overall model performance expressed by the mean RE (averaged over 2159
gridpoints) gives an indication how well, on average, the temperature and
precipitation fields are reconstructed over the whole area. The time series indicate
rather low RE values for the seasonal (AD 1500 to AD 1658) reconstructions,
though still positive.
This mainly reflects the low number of predictors available for this early period.
For the Mediterranean area there is only the precipitation predictor from Andalusia
(Southern Spain, Rodrigo et al. 2001) from 1501 onwards available. Therefore, the
averaged positive RE values for the early reconstructions are an indication that
remote predictors (from Central and Eastern Europe) can significantly contribute to
skillful temperature and precipitation reconstructions over the larger Mediterranean
area.
From the beginning of the 18 th century, with more temperature and precipitation
predictors, the quality of the monthly reconstructions increases. Provisional results
of time-dependent uncertainty ranges of the estimated temperature and precipitation
series reveal similar performance than of e.g. Briffa et al. (200 I a) which are based
on tree-ring data. The current level of uncertainty in the first centuries can be
reduced with the inclusion of further new early instrumental and proxy data from
areas sensitive to the Mediterranean area.
The spatial model performance gives insight in which areas reconstructions are
trustworthy. This has been demonstrated for the four winter extremes (Figures 4 and
5). Although the overall model performance for the wettest and driest winters
implies low reconstruction quality (not shown), there are regions within the
Mediterranean with skillful reconstructions. Therefore both, the overall and the
spatial model performance have to be considered when reconstructed temperature
and precipitation fields are climatologically interpreted.
A large fraction of precipitation and temperature variability can be explained by
large-scale circulation anomalies when averaging in space and time over adequately
large enough intervals (Trenberth 1990; Xu 1993; Qian et al. 2000; Xoplaki et al.