Low (web) Quality - BALTEX

11
Assessment of dynamical downscaling in river basins in Japan using the
Regional Atmospheric Modeling System (RAMS)
Koji Dairaku, Satoshi Iizuka, Wataru Sasaki, Roger A. Pielke Sr., and Adriana Beltrán-Przekurat
3-1 Tennodai, Tsukuba, Ibaraki 305-0006 Japan, e-mail: dairaku@bosai.go.jp
1. Introduction
The responses of the climate system to increases in carbon
dioxide concentrations and to changes in land use/land cover
and the subsequent impacts of climatic variability on
humans and natural ecosystems are of fundamental concern.
Because regional responses of surface hydrological and
biogeochemical changes are particularly complex, it is
necessary to add spatial resolution to accurately assess
critical interactions within the regional climate system for
climate change impacts assessments. We quantified the
confidence and the uncertainties of Type II dynamical
downscaling which the lateral and bottom boundary
conditions were obtained from Japanese 25-year ReAnalysis
(JRA-25) and assessed the value (skill) added by the
downscaling to a climate simulation in Japan.
2. Assessment of regional climate models
We investigated the reproducibility of present climate using
two regional climate models with 20 km horizontal grid
spacing, the atmosphere-biosphere-river coupling regional
climate model (NIED-RAMS) and the Meteorological
Research Institute Nonhydrostatic Model (MRI-NHM), both
of which used JRA-25 as boundary conditions. Two key
variables for impact studies, surface air temperature and
precipitation, were compared with the Japanese highresolution
surface observation, Automated Meteorological
Data Acquisition System (AMeDAS) on 78 river basins.
Results simulated by the two models were relatively in good
agreement with the observation on the basin scale. The
NIED-RAMS bias of 2 m air temperature (2mT) were less
than 0.5K and the bias of precipitation (P) were around 10%
in most of the river basins on annual averages for three years
(2002-2004). The biases over 29 years shown in the long
term experiment are similar to those of the three year
simulation (Figure 1). The model could add some
information as to where the larger scale information was
obtained.
3. Sensitivity to domain size and spectral nudging
scheme
A regional climate model often has sensitivity to model
configurations, such as domain size and nudging scheme.
We conducted sensitivity experiments to domain size and
nudging scheme using the NIED-RAMS. Smaller domain is
128x144 and larger domain is 216x240 (Figure 2). In each
domain, we conducted experiments with/without spectral
nudging scheme.
Spatial characteristics of the detected bias of 2mT on river
basins in Japan in the large domain were qualitatively
similar to that of the small domain. The model bias of 2mT
was quantitatively deteriorated in larger domain. On the
other hand, model bias of P was not significantly altered in
larger domain. The bias of P in June-July-August (JJA) was
comparatively strongly influenced by the domain size. It can
be attributed to the relatively weak synoptic-scale
disturbances in the summer season.
Spectral nudging scheme indicates some impacts on the
mean bias of surface variables (2mT and P) particularly in
JJA. But overall, the magnitudes of the impacts were not
significant. It can be speculated that a large part of the
domain area was dominated by the sea where sea surface
temperature forcing should play a significant role,
“nudging effect”, as a boundary condition.
Figure 1. The detected bias of 2m air temperature (°C)
(upper) and precipitation (%) (lower) in the NIED-RAMS
using station data, Automated Meteorological Data
Acquisition System, on 78 river basins in Japan for 29
years.
Figure 2. Experimental domains: small domain (left)
and large domain (right) in the NIED-RAMS.
Acknowledgements
This research was supported by the Global Environment
Research Project Fund (S-5, “Getting a feel for climate
change”) of the Ministry of the Environment, Japan.
Model bias detection program was provided by Prof.
Tanaka in Kyoto University.