Extreme Daily Precipitation in Germany
Jürgen Grieser
Download
Gumbel parameters here. Download
Visualization software here.
Download return periods and thresholds
here.
The German Meteorological Service (DWD) currently observes daily
precipitation sums at about 3000 stations. Records with not more than 1 missing
day are available at 1769 stations in Germany for the period 1951 to 2000.
Tim Staeger and myself fitted Gumbel distributions to the annual maxima
of daily, two-day and three-day precipitation sums. The results can be
downloaded as csv files here.
The files for the coefficient a and b have a different
number of columns. However, they all consist of a header line plus 6601 lines
of which each contains the parameter estimates and further information for a
grid point within Germany on 5’x5’ grid. The grid size is below 10km. We used
Kriging interpolation.
All files start with Longitude [°], Latitude [°], Altitude [m],
Gumbel_Parameter [mm], and Jackknife Error [mm].
Files with coefficients a of
the Gumbel Distributions furthermore provide the 2 columns Vertical Gradient
[mm/100m] and Explained Variance by Vert. Grad. [%].
Jackknife Error:
The Jackknife Error for each of the stations is the difference of the
observed or estimated value at the station and the interpolated value to that
station given the surrounding stations but leaving out the station for which
the Jackknife Error is to be estimated. We provide for each grid point the
Jackknife Error of the closest station. Thus, averaging over all the grid
points leads to an area-weighted Jackknife Error.
Vertical gradient of a:
We saw that the Gumbel coefficient a has a pronounced vertical
gradient. Therefore we performed linear regression with altitude for the 50
closest stations of each grid point. The resulting local vertical gradient and
its explained variance are provided in the according files.
Furthermore grids of return periods in years which correspond to
thresholds of 40, 50, 60, 70 80, 90, 100, 120, and 150 mm are calculated. Also thresholds
in mm corresponding to return periods of 2, 5, 10, 20, 50, 100, 200, 500, and
1000 years are provided. They can be downloaded here.
For further questions please drop a line to juergen.grieser@rms.com.
Latest Update: 30/10/2007