Geoscience Reference
In-Depth Information
As a minimum requirement, a yes/no answer is recommended to indicate
whether DQC has been applied or not. If the answer is positive, it would be good
practice to describe the degree of DQC applied to the data (e.g. subjected to logical
filters only; compared for internal coherency in sequence of observations, for spatial
consistency among suitable neighboring stations, for coherency with its
climatological values and limits) and to provide details on the employed techniques
and their application [AGU 03].
A simple example of inhomogenity in climate data series results from different
approaches for the computation of daily or monthly means of e.g. air temperature
from either observations at fixed times or from daily extremes of temperature (Table
1.1). The example shown in Table 1.2 is taken from the work of a new air
temperature map for the Greater Alpine Region (GAR) [HIE 09] using powerful
spatial modeling approaches including GIS techniques. However, before spatial
modeling could be started, station measurements had to be transformed to common
mean formula. Beside the formula for mean computation, the time reference system
used is also heterogenous within the GAR study region. It is obvious from this
simple example (which only tackles one out of several inhomogenities in climate
datasets) that DQC and data homogenization, in particular, are a laborious part of
climate modeling studies. Exclusion of this part of the modeling study could result
in systematic biases.
Figure 1.5.
Evolution through the year of the difference between various ways of calculating
daily mean temperature and 24-hourly observations average for the inner-alpine station
Puchberg in Austria, 1987-1996. Data source: Central Institute for Meteorology and
Geodynamics, Vienna, Austria (from [AGU 03])
Long-term series from measurements of automatic weather stations with hourly
values make it easy to compute the differences between various computation
formulas of daily means of air temperature used by NHMs. Selected examples of
differences between commonly used mean formulas and a 24-hourly mean are
shown in Figure 1.5. In fact, the widely used formula of (max+min)/2 show
differences of up to 1°C to the 24-hourly mean, which turns out to be larger than the
