Climate Variability
Principal Investigator
Hal Ritchie
Numerical Prediction Research Division - Montreal
Final Report: April 1, 1996
Data analysis software
A new improved ENM analysis software package has been develop by G. Brunet (RPN), B. Dugas (RPN)
and R. Wang (RPN). It is presently used for studying tropospheric NMC analyses, stratospheric
NMC analyses, SEF-AMIP and CCMA-GCMII-AMIP numerical integrations.
Diagnostics of low-frequency variability - AMIP analyses
Dr. Risheng Wang arrived in Montreal at the end of December and has begun his participation in the
project. His progress and integration into the research group and working environment are
very good. Together we are continuing preliminary diagnostics of the first SEF-AMIP
integration and the NMC analyses for the AMIP period. We are planning to present the
results of the analysis in Brunet et al. (1996) and Dugas et al. (1996).
In view of making diagnostics of the low-frequency with potential vorticity maps, G. Brunet,
S. Edouard and R. Vautard have prepared a three-dimensional climatology of potential
vorticity in isentropic coordinates and submitted the results to QJMRS, see Edouard et al.
(1996).
G. Brunet, R.Vautard and G. Plaut have done a preliminary analysis of the seasonal
predictions performed by H. Sheng. The approach was to compare the performance of the
SEF integration with two empirical models for monthly and seasonal forecast on the
Canadian region. A first draft of a technical report ( see Vautard et al.,1996) has been written
and should be made available in a few months. More details will be given in the course of the
coming fiscal year.
A predictability study using a shallow-water model has shown the advantage of using ENM
as a statistical basis for predicting the dynamics of the flow with an auto-regressive model for
each principal component. The results are presented in an article and have been accepted in
JAS under the condition of minor revisions.
The impact of numerical aspects on low-frequency variability in atmospheric
simulations.
Sensitivity to horizontal resolution, semi-Lagrangian and Eulerian schemes:
In order to assess the sensitivity of the low-frequency variability in seasonal forecasts to the
horizontal resolution in semi-Lagrangian and Eulerian schemes, northern winter (December,
January and February) simulations for ten years (the AMIP period, 1979-1988) have been
performed using various combinations of model resolution and formulations in the
operational Canadian global spectral forecast model. Comparisons have been performed
between semi-Lagrangian (SL) versions with linear unaliasing and Eulerian (EU) versions
with quadratic unaliasing running with triangular truncation on the same Gaussian grids,
namely: SLT31(EUT21), SLT63(EUT42) and SLT95(EUT63). The simulations are
assessed by comparing the forecasts to the corresponding analyses of mean states, including
mean sea level pressure, cross sections of temperature, zonal wind and meridional stream
function, as well as eddy meridional fluxes of momentum, heat and water vapor. The results
show that the latitudinal and geographic distributions of the sea level pressure improve with
increased horizontal resolution in the range of spectral truncation from SLT32(EUT21) to
SLT95(EUT63). Evaluating the mean of zonal wind one finds the positions of southern
hemisphere subtropical jet cores become more realistic with increased horizontal resolution
from SLT31 to SLT95. The value of jet centers is slightly stronger (about 2 metre/second)
than in the observations. One common deficiency from both the semi-Lagrangian scheme and
the Eulerian scheme is the stratospheric easterly bias (about 20 m/s). The cross section of
temperature shows the semi-Lagrangian scheme agrees favorably with observation at the
tropical tropopause, while discrepancies are present over the polar regions. Especially at the
south polar tropopause it is colder than observation and the Eulerian scheme. Various other
diagnostics related to dynmaics, moisture and heat budget variables are also being examined,
and are to be presented at the upcoming CMOS congress.
Improved numerical algorithms for stratospheric modelling:
One of the most evident deficiencies in the original RPN-AMIP simulation was the
inadequate representation of the stratosphere. In preparation for a more detailed examination
of the sensitivity of low-frequency variability to the vertical resolution and numerical
treatment of the stratosphere, modifications have been made to introduce a hybrid vertical
coordinate and to raise the model top to approximately the stratopause with an assocaited
increase in the number of vertical levels. In the new hybrid version all linear (semi-implicit)
vertical operators keep the same form as in the sigma version. The scheme is coded so that
all additional nonlinear terms arising from the coordinate transformation are treated explicitly
at the middle time level, which exploits the simplicity of a three-time-level discretization.
This formulation also removes a problem that has been diagnosed at the upper boundary of
the model. By examining the thermodynamic equation it can be seen that the conventional
imposition of a zero vertical velocity on the upper sigma boundary forces the mountains to be
reflected in the temperature field at the model top when the trajectories pass over the
mountains. This can be seen, for example, over the Himalayas in forecasts of the
temperature at 1 hPa with a sigma version of the model, whereas no evidence of this
pathological behaviour is present in the corresponding forecast with the hybrid version of the
model. Testing of a version with complete physical parameterizations has begun, and work
is in progress to redo the RPN-AMIP simulation with the hybrid coordinate version of the
model. Associated presentations have been submitted for the upcoming CMOS congress.
Reference