![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
The best ensembles of RCMs for climate change projections in Ukraine |
VerfasserIn |
Svitlana Krakovska, Natalia Gnatiuk, Liudmyla Palamarchuk, Iryna Shedemenko |
Konferenz |
EGU General Assembly 2013
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 15 (2013) |
Datensatznummer |
250072542
|
|
|
|
Zusammenfassung |
Application of climate models ensembles instead of individual ones became very useful and
prominent technique in modern climate change research. But methodology of models’
selection to the ensemble should be different in dependence of the purposes of its application
and be based on certain criteria. The proposed research aimed to obtain the best ensembles
for climate change projections in Ukraine in the 21st century for two key climatic
characteristics: air temperature and precipitation amount. With this purpose 14 RCMs from
the European project FP-6 ENSEMBLES scenario SRES A1B and their various ensembles
were verified for the periods 1961-1990 and 1991-2010 against the data from the
E-Obs database in every 25x25 km grid box and averaged over the territory of the
country.
The main criteria for RCM selection to the ensembles for monthly air temperature were
following statistical characteristics: multi-year annual mean and its standard deviation as
amplitude of the annual variation, root mean squared differences (RMSD) and absolute
maximum and minimum errors. Coefficients of correlation were calculated too, but they are
rather not indicative for model successfulness in this case since almost all RCMs have the
coefficients higher than 0.99 demonstrating that annual course of temperature is perfectly
simulated by them. Based on the above criteria 4 RCMs were excluded from the ensemble
due to very high max and min absolute errors (over ±4oC) in both control periods. Therefore
the ensemble of 10 RCMs has been defined as optimal for air temperature projections in
Ukraine.
One more method was applied for RCMs ensembles selection and verification. The main
idea is to use the same methodology as proposed for climate projection obtaining.
Specifically to build a projection for period 1991-2010 from 1961-1990 based on the same
RCM ensembles and compare with data of E-Obs for this modern time. The criteria of
“successfulness” were the same as above and for the defined optimal ensemble of
10RCMs all errors were minimal specifically an average areal absolute error was only
-0.07oC.
The same methodology was applied for multi-year monthly precipitation amount. But in
this case correlation coefficients should be one of the most decisive parameters. Another two
are RMSD and standard deviation, and Taylor diagrams are most useful for precipitation
verification in this case. Unfortunately just 8 RCMs in the period 1961-1990 and 5 RCMs in
period 1991-2010 had positive correlation coefficients and just 4 RCMs had them higher than
0.65. That is why possibility to combine an ensemble was much limited than for
temperature. Nevertheless the same methodology as for temperature to build projections
for precipitation from past to modern period was applied for the ensemble of 4
RCMs. And an obtained mean areal correlation coefficient between precipitation
projection of 4RCMs ensemble and E-Obs data for period 1991-2010 was 0.69, but for
averaged over territory of Ukraine multi-year monthly precipitation annual course
as high as 0.87 and RMSD was just 6.7 mm that is indeed good result. Thus the
ensemble of 4 RCMs has been defined as optimal for precipitation projections in
Ukraine.
And the last conclusion from the RCMs verification is that RCM REMO (MPI-M,
Hamburg) had the best statistical characteristics both for temperature and precipitation in
Ukraine. Therefore this RCM could be recommended as the best one for those investigations
where application of ensemble is impossible or useless and daily data are needed
(hydrological, agricultural or other applications). |
|
|
|
|
|