Geoscience Reference
In-Depth Information
A very detailed study on climate data homogenity is available for the Greater
Alpine region from several research projects (e.g. [AUE 07]). Some important
results of these studies are summarized in Table 1.3 showing that data inhomogenity
is immanent to climate data studies even on a very short time scale. The number of
detected breaks is quite high and the mean length of homogenous sub-interval is in
the range of 10-30 years for all climate variables shown. Results in this table are
derived from selected monthly data series covering monthly means and monthly
sums. If, however, climate extremes or daily data series are studied, the problem of
data homogenity is even more pronounced.
1.2.2.
Methods for climate DQC
DQC is part of the core of the whole data-flow process. In fact, it has to ensure
that data are checked and is as error-free as possible. All erroneous data have to be
eliminated and, if possible, should be replaced by corrected values (while retaining
the original values in the database).
Useful tools of DQC for climate data are (Aguilar
et al
., 2003):
a)
Gross error checking
: report what kind of logical filters have been utilized to
detect and flag obviously erroneous values (e.g. anomalous values, shift in commas,
negative precipitation, etc).
b)
Tolerance test
: documents to which tests have been applied, to flag those
values considered as outliers with respect to their own climate-defined upper/lower
limits. The tests provide the percentage of values flagged and the information on the
approximate climate limits established for each inspected element.
c)
Internal consistency check
: indicate whether data have undergone inspection
for coherency between associated elements within each record (e.g. maximum
temperature < minimum temperature; or psychrometric measurements, dry-bulb
temperature wet-bulb temperature).
d)
Temporal coherency
: inform if any test has been performed to detect whether
the observed values are consistent with the amount of change that might be expected
in an element in any time interval and to assess the sign shift from one observation
to the next.
e)
Spatial coherency
: notify if any test is used to determine whether every
observation is consistent with those taken at the same time in neighboring stations
affected by similar climatic influences.
Figure 1.8 shows the results from a detailed homogenization study of climate
time series for the GAR, which also included estimation of outliers and gap filling.
Whereas the time series of outlier rates (¿gures on the left) indicate more about
internal system stability of meteorological networks the gap rates (¿gures on the
right) seem to react more to external inÀuences. It is interesting to see from Figure
1.8 that both outliers and gaps increased since the 1980s, which was the beginning
of automation of climate networks in the study region.
