Variability Analysis of Surface Climate Observations
VASClimO was a joint climate research project of the Global
Precipitation Climatology Centre (GPCC) at
the German Met Service (DWD) and the Institute
for Atmosphere and Environment – Working Group for Climatology
at the Johann Wolfgang Goethe University Frankfurt.
The project was funded by the Bundesministerium für Bildung,
Wissenschaft, Forschung und Technologie BMBF within the German Climate Research Program DEKLIM(DEKLIM
Project-No.: 33 11 0307 - "Development of an observational data basis
(Europe and global) for DEKLIM and related statistical analysis with regard to
climate variability on a decadal to centennial time scale"). Funding was
provided over a 5-year period for 3 scientists. Two scientists (Dr. Christoph
Beck and myself) spent their time in part project A at GPCC. Their task was the
creation of worldwide gridded datasets of 4 meteorological variables
(precipitation, near surface temperature, snow cover and air pressure) in 2
temporal resolutions (monthly and daily). The datasets should have been built
on long, quality controlled (homogeneity, outliers, station location) records
of observations.
The leader of project part A (Dr. Bruno Rudolf) refused to provide
programmable computers. The project scientists where allowed to use 1.5Gb of
disk space on the big computer of the German Met Service. This was the first
measure of Bruno Rudolf that ensured that no product at all could be generated
within the project.
Dr. Bruno Rudolf forced the project scientists to use the Lademodul. The
Lademodul is a computer software programmed by Jürgen Lang’s company
MeteoSolutions. The software puts data into the DWD database and by that is
supposed to do a station metadata control. Therefore it should have been used
for the merging of data from different sources. The DWD-senior scientist Udo
Schneider takes responsibility for the design of this software.
The project scientists showed clearly and early that the use of this
software ensures that the meta-data quality of datasets is decreased (there
were cases where the data had to be either attributed to wrong stations or
“new” stations had to be wrongly “created”). Important and simple tests
(station in right country? station over sea? Other stations available with the
same record of observations) were not part of the quality control strategy with
the Lademodul. Strange enough, the “quality control” with the Lademodul which
only checks station meta data had to be done for each observation instead of
each station (i.e. about 600 times more often than necessary). This and the
unhandy input format increased the workload considerably.
I do not further report on the strange features of the Lademodul, though
there are plenty of disadvantages to be listed. I will report on that in
another frame. It took me 2 years of the project until I finally refused to
further use the Lademodul. These 2 years caused enough damage to the data.
A more prominent point is that Dr. Bruno Rudolf gave instructions to do
senseless work. Here comes a nice example: We got a data delivery of monthly
temperatures at stations with WMO number. Since the stations are uniquely
identified by that number no further meta information was provided. Dr. Bruno
Rudolf insisted on the use of the Lademodul and forced me to invent station
metadata (coordinates, barometer height, altitude, country code, and others) to
perform the whole Lademodul data quality control. This is ridiculous, to put it
into polite words.
There is a bunch of reasons:
First, one has to have an income.
Quitting a job in Germany without an alternative means having at least 3 months
no income. Afterwards the unemployment insurance jumps in for 9 months.
Second, Frankfurt is my hometown and
I didn’t want to leave my familiar environment again if I was not completely
sure that there was no chance of a happy end.
Third, Bruno Rudolf promised that he
takes care of his employees. And that he takes responsibility if we follow his
commands closely instead of doing reasonable and project relevant work.
Forth, the leader of the whole
project (Prof. Schönwiese) asked me to stay in the project.
We did complain. First the project
leader Prof. Schönwiese was informed about part of the senseless instructions.
He asked me to stay and try to do the best out of it. The division director
Volker Vent-Schmidt was informed and threatened with “actions” in case I
further inform the project leader. Later, when the GPCC changed division the
other director Dr. Helene Bartels was informed and asked me to just not take
Dr. Bruno Rudolf serious.
Finally - one should wash dirty
washings at home - we did not inform the ministry funding the project and
showed the outside world full loyalty to GPCC and DWD on meetings and
conferences.
What was the outcome?
Nothing. No product could be created
within the project. With the years passing by it became more and more clear
that even the work instructions and information that sounded reasonable in the
first place was wrong and ensured that we spent most of our project time not
with research and development of products but with mixing data and ensuring
that their quality was not enhanced.
How did we treat this?
We had the choice to give up
completely or to try even harder. We went for the second. Dr. Christoph Beck
managed to get some disk space at a university not related at all to the
activity and I bought a private laptop and brought it to office every day. Now
we could work. Nobody could restrict it - we thought. The result was that Bruno
Rudolf refused to provide important data (like long-term averages of which he
states (and published) that they exist within GPCC for 30,000 stations).
Parallel to this strategy he gave new tasks that ensured that we could not do our
basic work. He did not allow me to publish my analysis of different
interpolation strategies (resulting in the reasons for the one I finally used)
but forced me to be a lead author of a paper on Köppen climatology though my
knowledge on this subject was limited. I tried hard to convince him that this
may not be the best idea. However, I was declared lead author, we came up with
a manuscript and it was recaptured by colleagues from Vienna who started the
whole activity in the first place.
Nevertheless, we used the data we
had and the non-project computers plus a considerable amount of spare time,
performed homogeneity tests (and homogenisation if possible), removed outliers,
investigated station density and interpolated monthly mean precipitation time series
in order to provide at least one product to the project. We offered the results
of this mainly private activity to the part project leader Dr. Bruno Rudolf who
refused to accept it unless it is done a second time. The amount of about 5000
stations worldwide had to be increased by about 1000 French stations. This kept
us busy again, enhanced the global data availability by 20% and introduced a
spatial data inhomogeneity. After we finished this work (again on private
computers and in spare time in the hope to provide something at the end of the
project) he forced us to do everything again and to also include the about 3300
German stations (i.e. to enhance station density in a region where station
density is high anyway). Now the final product consists of 9343 stations worldwide (of which nearly
half are in France and Germany).
All this was very suspicious and I
asked Bruno Rudolf whether his promise to take responsibility for the project
would still hold. After the disaster with the years of meaningless “quality
control” and his clear No to any technical support (neither
hardware nor software) I wondered how he would take responsibility for that. So
I made it very clear again that I was not prepared to pay the bill for his
decisions especially since I always informed him of what I knew would be the
results if we did what he want. He confirmed that he would take the
responsibility and I should hand over the one and only product to the project:
A quality controlled (we meanwhile did a quality control of the observations,
homogenized and removed outliers) 50-year monthly precipitation climatology for
the land surfaces of the world (excluding Antarctica and Greenland). Needless
to say that the product would have been better if Dr. Bruno Rudolf did not
refuse to provide the GPCC long-term averages and if all available data could
have been used and not only the ones which went through the Lademodul. And if
the project scientists were not forced by Dr. Bruno Rudolf to waste time with
complete useless activities there would have been even time to provide a second
or more of the 8 promised products.
As a consequence Dr. Bruno Rudolf
suggested that I should not even try to work as a meteorologist again. A year
later he became a division director. The colleagues wrote a final report in
which the success of the project is highlighted though only one of 8 products
is available. I had no job anymore and left country after I had spent money and
spare time and a lot of energy and commitment in my work. And until the publication
of the final report I kept the real activities in the project secret.
Needless to say that there is no
scientific publication at all of the “work” performed by the 2 project
scientists of VASClimO part A (10 man years).
The Final Dataset
Here are the gridded data I produced
following Christoph Becks homogeneity and outlier tests of the incoming data.
They are not pure interpolations of observations as it is usually done by GPCC
(with Shepards method using 20 neighbours). Deviations from the long-term
averages are interpolated. This ensures that more local information is used due
to the large number of stations for which averages are available. This strategy
is known as Climatology Aided Interpolation (CAI). However, the 27,000
available averages are taken from the FAO database (completely uncontrolled)
since the part-project leader and head of the GPCC, Dr. Bruno Rudolf, refused
to provide the averages which he pretend to have for 30,000 GPCC stations.
The final dataset consists of
gridded monthly data for the global land areas (excluding Greenland and
Antarctica) from 1951 to 2000. It is gridded at three different resolutions
(0.5° lat/lon, 1.0° lat/lon, 2.5° lat/lon) on the basis of long and nearly
gap-free quality controlled station records. The dataset is now available for
use in scientific research and can be downloaded as version 1.1 dated August
2005.
Downloads:
Readme with file format details.
Resulting spatial statistics (averages, standard deviations,
coefficients of variance, absolute and relative linear trends, trend-noise
ratio, and Mann-Kendall significance):
Statistics 0.5°
resolution (8Mb)
Statistics 1.0°
resolution (2.2Mb)
Statistics 2.5°
resolution (.4Mb)
Zipped
video clips (avi):
January, February,
March, April,
May, June, July, August, September, October,
November, December,
interpolated annual average, annual average of monthly interpolations
Animated Gifs
January, February,
March, April,
May, June, July, August, September, October,
November, December,
interpolated annual Average, annual average of monthly interpolations
Visualisations:
Maps of resulting statistics, error
estimates and trends.
Time Series of Global Means of
Monthly Mean Daily
Precipitation
Annual Means of Monthly
Precipitation Sums
Please cite the dataset as
Beck, C., J. Grieser and B. Rudolf (2005): A New Monthly Precipitation
Climatology for the Global Land Areas for the Period 1951 to 2000. Published in
Climate Status Report 2004, pp. 181 - 190, German Weather Service, Offenbach,
Germany. [pdf]
Further Reading
Grieser, J., and C. Beck, 2006: Variability and Triggering Factors of
Observed Global Mean Land-Surface Precipitation since 1951. DWD,
Klimastatusbericht KSB 2005, accepted. [pdf]
