Figure 2 -

Global high resolution climate reconstructions
Martina Schubert-Frisius (1), Frauke Feser (1), Matthias Zahn (1), Hans von Storch (1), and Sebastian Rast (2),
(1) Helmholtz-Zentrum Geesthacht (HZG), Zentrum für Material - und Küstenforschung; Germany, (2) Max-Planck-Institut für Meteorologie Hamburg (MPI-M Hamburg); Germany
Experiments and results:
Sensitivity of the nudging parameters:
Long-term reanalysis data already exist in coarse resolution. However, for
advanced climate studies downscaling is needed. In a global high resolution
atmospheric model the downscaling will be performed by using the spectral
nudging technique. This kind of data assimilation conserves large scale
structures but provides in addition details at a finer resolution.
For the results presented here we have selected the nudging configuration types N1, av and a nudging strength of
0.5 up to the wavenumber 42. Only the divergence and vorticity have been used as nudging variables for the
assimilation run (year 2004). Figure 3 compares the temperature of the simulation with and without nudging. The
improvement in the representation of NCEP1 large scales features becomes evident.
The correlation of vorticity between ECHAM6 and NCEP1 (Fig. 7) is sensitive to the nudging
wavenumber range (T42,T21,T11 and T05). The part of the spectrum where nudging takes place
become clearly evident by high values near one. Outside of the part much smaller correlation
values arise. The zonally averaged correlation of temperature fields between ECHAM6 and
ERAINTERIM has a minimum near the equator. There, the RMSE also becomes minimal. A
drastic decrease of correlation and increase of RMSE can be observed for the spectral
wavenumber ranges T11 and T05.
Model, Reanalysis Data and Spectral Nudging
The GCM ECHAM6 (T255L95) developed at the MPI-M Hamburg (Germany)
is used. The model fields will be nudged by NCEP1 reanalysis data
(T62L28). Selected wavelengths of more than 1000km for the variables
vorticity, divergence, temperature and the logarithm of surface pressure is
imposed onto the simulated GCM counterparts at levels above 750 hPa. SST
and sea ice fields are taken from the NCEP1 data set.
Different nudging configurations are applied to find out the best agreement
between the new produced reanalysis and observational data: Mainly, we
distinguish between 3 types of horizontal nudging profiles (see Fig. 1).
Figure 3: Temperature at the lowest model level for 31.01.2004, 00UTC. Left panel: NCEP1-reanalysis, middle panel: difference between ECHAM6 with
spectral nudging and NCEP1, right panel: difference between ECHAM6 without spectral nudging and NCEP1.
Figure 7: The correlation in spectral space between the nudged ECHAM6 model fields of the vorticity at level 82 (~500hPa) and the
corresponding NCEP1 data for different nudging wavenumber ranges (T42,T21,T11 and T05).
The time series of the temperature at 2m height and the wind speed at 10m height at selected locations display a
good agreement with observed DWD station data (see Fig. 4). The values of the bias, the time correlations and
the brier skill score demonstrate an improved conformity between these two data sets in the small scaled
meteorological phenomena. The brier skill score employs NCEP1 as the standard hindcast.
Figure 8: Zonally averaged correlation (red lines) and RMSE (blue lines) between the nudged ECHAM6 model fields of the temperature at
850 hPa and the corresponding ERAINTERIM data for different nudging wavenumber ranges (T42,T21,T11 and T05).
Figure 1: Different types for spectral nudging: a) zonal, type N0, b) zonal = meridional, type N1 and c)
rhomboidal, type N4. In the figures m, l and n denote zonal, meridional and total wavenumber, respectively.
Figure 2 shows 3 vertical nudging profiles with their strengths which are
chosen for our sensitivity studies.
Figure 4: Time series of the nudged ECHAM6 and observed station data (DWD) for 2m temperature (left panel) and 10 m wind speed (middle panel) at
Hamburg-Fuhlsbüttel (Germany). Right panel: Bias, time correlation and brier skill score at selected DWD stations for 2m temperature and 10 m wind speed.
Figure 5 shows time correlations between ECHAM6 and reanalysis data for various quantities. High values can
be found in the extra tropics and less satisfying results in the tropics, where the variance is relatively small.
Figure 2: Different types for the vertical nudging profile: a) vertically constant (echam6 default), b) plateau
(dvw) and c) parabel (av).
 G  X NCEP
 X   Fm  X
X represents the model variable, XNCEP1 the NCEP1 data, Fm the model
tendency and G (XNCEP1 – X) is the product of the nudging coefficient G (s-1)
and the difference between the variable values of NCEP1 and the model.
Figure 5: Time correlations between ECHAM6 nudged and ERAINTERIM for 2m temperature (left panel), for 10m wind (middle panel) and between ECHAM6
and NCEP1 for the daily sums of precipitation (right panel).
The spectra of kinetic energy at 500hPa, Fig. 6, reveal nearly
identical results for total wavenumbers n ≤ 42. For higher wavenumbers ECHAM6 has larger amount of kinetic energy. The
inertial sub-range split into low wavenumber part (~ n-5/3) and
higher wavenumber part (~ n-3).
Acknowlegement:This research is funded by the german science foundation (DFG) within the Clustre of Excellence Climate System Analysis and Prediction (CliSAP).
Figure 9: Comparison of the cyclone tracks for typhoon Tokage
(october, 2004) in the various nudged ECHAM6 simulations
(T42,T21,T11 and T05) and best track data (crosses). The
inserted diagrams display the time evolution of distance to best
track, wind speed and MSLP.
The nudging technique attracts the model trajectory toward reanalysis data
by adding a relaxation term to the model equations:
Numerous test simulations with different
combined nudging configurations were
performed to achieve the best consensus
between the observed typhoons (best track
data and their calculated counterparts with
regard to their positions, wind speed and
sea level pressure (Fig. 9). For nudging
range T42 and T21 ECHAM6 delivers
satisfactory results in simulating the track of
typhoon Tokage. However their trends of the
life cycles of wind speed and MSPL show
distinct weaker tendencies.
Figure 6:
energy spectra for
ECHAM6 (light blue)
and NCEP1(blue).
• The spectral nudging technique is an economic and effective method for data assimilation.
• The good accordance in near surface meteorological quantities of nudged ECHAM6 model
fields, observed station data and reanalysis data respectively can be confirmed using the
statistical methods. An exception is found in the tropics with low correlations.
• The quality of the new created global reanalysis data reveals a strong dependency on the used
nudged wavenumber range.
Kalnay et al.,The NCEP/NCAR 40-year reanalysis project, Bull. Amer. Meteor. Soc., 77, 437-470, 1996.
Dee DP et. al., 2011: The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q. J. R. Meteorol. Soc.
137: 553–597.
Stevens, B., Giorgetta, M. A., Esch, M., Mauritsen, T., Crueger, T., Rast, S., Salzmann, M., Schmidt, H., Bader, J., Block, K., Brokopf, R.,
Fast, I., Kinne, S., Kornblueh, L., Lohmann, U., Pincus, R., Reichler, T., & Roeckner, E. (submitted). The atmospheric component of the
MPI-M Earth System Model: ECHAM6.
Centrum für Erdsystemforschung und Nachhaltigkeit (CEN)

similar documents