Item

Abstract

In the European Commission (EC) project "Regionalization of Anthropogenic Climate Change
Simulations, RACCS, recently terminated, 11 European institutions have carried out tests of
dynamical and statistical regionalization techniques. The outcome of the "dynamical part" of
the project, utilizing a series of high resolution LAMs and a variable resolution global model
(all of which we shall refer to as RCMs, Regional Climate Models), is presented here. The per-
formance of the dqterent LAMs had first, in a preceding EC project, been tested with "perfect"
boundary forcing fields (ECMWF analyses) and also multi-year present-day climate simula-
tions with AMIP "perfect ocean " or mixed layer ocean GCM boundary conditions had been
validated against available climatological data. The present report involves results of vali-
dation and analysis of RCM present-day climate simulations and anthropogenic climate
change experiments. Multi-year (5 - 30 years) present-day climate simulations have been per-
formed with resolutions between 19 and 70 km (grid lengths) and with boundary conditions
from the newest CGCM simulations. The climate change experiments involve various 2xCO2 -
]xCO2 transient greenhouse gas experiments and in one case also changing sulphur aerosols.
A common validation and inter-comparison was made at the coordinating institution, MPIfor
Meteorology. The validation of the present-day climate simulations shows the importance of
systematic errors in the low level general circulation. Such errors seem to induce large errors
in precipitation and surface air temperature in the RCMs as well as in the CGCMs providing
boundary conditions. Over Europe the field of systematic errors in the mean sea level pressure
(MSLP) usually involve an area of too low pressure, often in the form of an east-west trough
across Europe with too high pressure to the north and south. New storm-track analyses confirm
that the areas of too low pressure are caused by enhanced cyclonic activity and similarly that
the areas of too high pressure are caused by reduced such activity. The precise location and
strength of the extremes in the MSLP error field seems to be dependent on the physical param-
eterization package used. In model pairs sharing the same package the area of too low pressure
is deepened further in the RCM compared to the corresponding CGCM, indicating an increase
of the excessive cyclonic activity with increasing resolution. From the experiments performed
it seems not possible to decide to what extent the systematic errors in the general circulation
are the result of local errors in the physical parameterization schemes or remote errors trans-
mitted to the European region via the boundary conditions. Additional errors in precipitation
and temperature seems to be due to direct local effects of errors in certain parameterization
schemes and errors in the SSTs taken from the CGCMs. For all seasons many biases are fOund
to be statistically significant compared to estimates of the internal model variability of the time-
slice mean values. In the climate change experiments statistically significant European mean
temperature changes which are large compared to the corresponding biases are found. How-
ever, the changes in the deviations from the European mean temperature as well as the changes
in precipitation are only partly sign wcan

ce and are of the same order of magnitude or smaller
than the corresponding biases found in the present-day climate simulations. Cases of an inter-
action between the systematic model errors and the radiative forcing show that generally the
errors are not canceling out when the changes are computed. Therefore, reliable regional cli-
mate changes can only be achieved after model improvements which reduce their systematic
errors sufficiently. Also in future RCM experiments sujiciently long time-slices must be used
in order to obtain statistically sign ijicant climate changes on the sub-continental scale aimed
at with the present regionalization technique.