DAILY AIR TEMPERATURE AND PRESSURE SERIES ... - BALTEX

204 ANDERS MOBERG ET AL.
5.6. DISCUSSION OF THE SEA LEVEL PRESSURE SERIES
We have reconstructed a series of daily mean sea level air pressures for Stockholm
from the original observational data. Performing the reductions to sea level before
1859 was the most difficult part of the work, because little is known about the
properties of the used barometers and their positions. Furthermore, no barometer
temperatures were observed before 1785, so these values had to be estimated
from outdoor temperatures. Comparisons with data from other European stations
reveal that the air pressure values for Stockholm, as calculated directly from the
observations, exhibit several sub-periods with various systematic biases. A homogenization
was therefore done by adding certain corrections obtained from
homogeneity tests. The homogenization is expected to correct for most of the
systematic errors in the original data, and also for possible systematic errors introduced
by us whenever we made wrong assumptions concering the properties of
the early barometers and conditions around the observations.
A plot (Figure 11b) of the homogenized annual mean pressures back to 1756
for both Stockholm and Uppsala (Bergström and Moberg, 2002) shows that the
year-to-year variability is very similar in these two series. The similarity is even
more striking in plots of daily air pressures. Two examples of time-series plots of
daily pressures in selected years (1788 and 1819) are shown in Figure 13. The plots
illustrate that the day-to-day pressure changes is almost identical in the two series.
This is to be expected for two such nearby stations. Other selections of years (not
shown) reveal the same high degree of similarity. This is a very good indicator that
both records are very reliable reconstructions of daily air pressures. In particular,
the data can be used with a high degree of credibility as input for drawing pressure
charts over Europe for any particular day.
Regarding long-term trends (multi-decadal scales), the reliability of the data
must be considered as lower than for the conditions in individual days. This is
rather obvious because the pressure difference between various multi-decade long
periods is small in comparison with many of the errors in the observed data.
Furthermore, the homogenization of the long-term trends was largely based on
comparison with data for surrounding stations, which means that the reliability of
trends is dependent on the quality of the reference data.
The reliability of trends is regarded as very high back to 1890, where the homogenization
was based on comparisons with the NACD data set of Frich et al. (1996).
This data set consists of a large number of series which have been homogeneity
tested and maintained by the national meteorological services in several countries.
Although we used homogenized reference data back to 1780, the reliability of
pressure trends before 1890 must be expected to decrease further back in time.
This is because the reference data we used (from Jones et al., 1999) had, in their
turn, been homogenized based on comparisons with fewer and fewer surrounding
stations further and further back in time. Finally, the Stockholm pressure series
can not be used at all for studies of changes in pressure trends before 1780, as the