Geoscience Reference
In-Depth Information
[AGU 03]) that various homogenization tools work well for monthly time series, but
homogenization of daily series or series with even higher temporal resolution is still
not solved properly.
Figure 1.10.
Inhomogenities in monthly data series of air pressure, cloudiness, air
temperature, precipitation and sunshine for the GAR (from [AUE 07]). Black bold
lines show the mean of inhomogenity series, which implies that even
averaging over many series cannot level out inhomogenities
As previously outlined climate time series homogenization includes two
important steps: first, the detection of breaks in the series (which need the creation
of a reference time series), and second, data adjustment. Homogenization should be
based on available metadata, which means that breakpoints should, as much as
possible, be reflected by metadata information. However, many inhomogenities of
climate time series are not captured by metadata (only perfect metadata would
include all breaks). If available, parallel measurements from sensors, station
locations, etc, should be used for adjustments. Today all homogenization tools for
breakpoint detection use relative homogenity tests, which mean that they detect
