final_program_abstracts[1]

11 IMSC Session Program
Using observations to verify climate predictions and the
evaluation of models used at Hungarian Meteorological
Service through statistical methods
Friday - Poster Session 7
Peter Szabo
Observational studies showed that changes in precipitation and temperature are
amplified at the tails of the probabilty distribution function. One of our objective is to
show whether there have been remarkable changes in the frequency of observed
extreme events through Hungarian observations available from 1961, and also while
focusing on the extremes is to compare our model outputs (ALADIN-Climate and
REMO) with this gridded data set. The model outputs have also been analyzed for the
projections: we have time slices in case of ALADIN-Climate (from 1961 to 2100) and
the REMO has a transient run for the period of 1951-2100.
The ALADIN-Climate model was developed by Meteo France (based on the short
range numerical weather prediction model ALADIN) and has both 10 km and 25 km
of horizontal resolution (the latter is used only for the experiments for the past). The
REMO model developed by Max Planck Institute for Meteorology is available on 25
km horizontal grids. The validation experiments are performed with perfect lateral
boundary forcings (ERA-40), or driven by GCMs (ARPEGE-Climat for ALADIN-
Climate and ECHAM for REMO). In this study we analyze several climate extreme
indices (included complex indices) for the Carpathian Basin. We try to determine the
spread and the average of the model outputs, though they differ in resolution, forcings
and in terms of the observations as well.
The results are complemented with significance tests using t-test, Welch-test and
Mann-Kendall-test: these indicate that there have been a trend in the past events or
whether the model deviations from observed climate and with respect to the future
simulations are considered to be significant or not. However, it requires more
sophisticated mathematical analysis, a bivariate distribution is the most common
method for modeling extreme events.
Abstracts 327