How to move the gravitywave parameterization problem forward Some thoughts

1
How to move the gravity-wave parameterization
problem forward? Some thoughts
NCAR TIIMES Gravity-Wave Retreat, 2006

• Ted Shepherd
• Department of Physics
• University of Toronto

2
Observations
3

• The first principles approach (Tim)
• Need to test parameterizations against highly
  resolved simulations
• Need to get reasonable results with reasonable
  parameters
• Need to know that sensitivity to climate
  perturbations is realistic
• Filtering effects are probably robust
• Source changes are more of a challenge
• The users approach.

4

• We need to identify the aspects of GW
  parameterizations that matter the most for
  weather and climate simulation
• Despite the number of different parameterizations
  and the intensity of debate between some of
  their proponents the choice of parameterization
  seems not to matter greatly
• However most comparisons have not been
  well-constrained

5

• Once the source spectrum is constrained, the only
  important parameter seems to be intermittency
  (McLandress Scinocca 2005 JAS)
• Determines the breaking height
• This is not very surprising, in light of
  downward control
• Some assumptions
• CLs always absorb
• CLs always reflect
• No horizontal propagation

6

• The partitioning between nonlinear drag and
  critical-level drag depends on the scheme, but
  the net drag is the same

From McLandress Scinocca (2005)
7
In an active GCM such as CMAM, one can actually
rely on critical-layer drag alone! But if the
drag only sees the zonal-mean wind, then one
needs nonlinear drag

8

• Instantaneous snapshot of SKYHI
• zonal winds for various altitudes,
• during a model July
• Increasing gravity-wave activity
• with increasing altitude

From Koshyk et al. (1999 JGR)
9
Horizontal wavenumber spectra (n spherical
harmonic index) of kinetic energy for SKYHI and
CMAM Straight lines show -3 and -5/3
slopes Charney-Drazin filtering is
evident Shallow spectra emerge with increasing
altitude Figure courtesy of John Koshyk
10

• This all confirms the suspicion that many have
  had for a long time that the key issue is the
  source spectrum, and perhaps to a lesser extent
  intermittency
• However it is also essential that
  parameterizations are implemented in a
  momentum-conserving way, and that there is no
  Rayleigh drag or zonal-mean sponge layer (Shaw
  Shepherd JAS, in press) otherwise robustness is
  lost
• Remarkably, this is far from the case with
  climate models!

11

• From the modelling side, we need to identify
  where it is that gravity wave parameterization is
  most important (for either climate or climate
  change), and assess the robustness of different
  model results in this respect
• Polar vortex, esp. in the SH
• Summer mesopause
• MLT more generally difficult
• Tropical upwelling
• SAO and QBO difficult

12

• First we need to identify the principal
  climate-change uncertainties associated with
  gravity-wave drag (insisting on momentum
  conservation!)
• SPARC CCMVal is a good framework for this
• And then we need to develop a better
  understanding of the sensitivities in
  well-constrained comparisons
• Emerging SPARC initiative on dynamical processes
  should provide a good framework

13

• Development of physically-based source
  parameterizations (which respond to climate
  change) is a very positive step
• We need to assess their sensitivity (e.g. to
  climate variations) and identify their role in
  the simulations
• Impact of changes in sources vs impact of changes
  in GW filtering
• Its not obvious that a predicted change in the
  source is better than no change!

14

• In the extratropics, middle atmosphere data
  assimilation should provide very useful
  constraints on GWD
• Planetary waves and zonal winds in the
  troposphere and stratosphere should be about
  right, hence the filtering of GW fluxes
• This will also slave the large-scale mesospheric
  fields to a large extent
• Increments from temperature observations will
  likely mainly reflect errors in GWD

15

• In principle, if GW parameters are a control
  variable in the data assimilation, then they can
  be constrained by the temperature observations
• This is a developing theme within the SPARC Data
  Assimilation Working Group
• Can one use an instrument forward model
  off-line, acting on the parameterized GW
  spectrum, to predict GW variances observed by
  satellites?
• Would likely need to be statistical

16

• The large volume of satellite data relevant to
  GWs raises the prospect of performing reasonable
  statistical tests
• Intermittency
• Sensitivity of sources to local conditions
• But how much of a constraint do these satellite
  observations place on the part of the GW spectrum
  that matters?
• Need to relate what the satellites measure to
  the full spectrum (via mesoscale models, field
  experiments), and use the satellites to
  extrapolate to global fields