Back to CAM Development Plan

Back to AMWG Homepage

DYNAMICAL CORE EVALUATION

Global baroclinic core evaluation for CAM candidates (The following is derived from the Spring 2006 AMWG Meeting and is intended as a starting point for discussion. Send comments, criticisms, suggestions etc. to Dave Williamson (wmson@ucar.edu) for compilation)

As usual, assume that the basic numerical approximations have demonstrated desired characteristics in Cartesian geometry. The additional tests here are intended to evaluate the scheme in spherical coordinates specifically for atmospheric model application.

Required properties for a global dynamical core:

• Global energy conservation (most likely via a fixer applied every time step)
• Reasonable average KE spectrum, KE loss through diffusive aspects ~2Wm^-2 (both spectrum and loss with complete parameterization suite such as in APE test below)

Desired properties:

• Locally conservative
• Consistent atmospheric mass and tracer conservation
• Shape preserving

Unfortunately much evaluation remains somewhat subjective. In many cases there are no absolute criteria and we are looking for potential flaws in a core as well as strong points. The tests are intended to identify schemes for which it is worth taking the final step of incorporating the complete CAM parameterization suite, including land and sea-ice for a standard climate simulation test and consideration to become the standard CAM dynamical core.

---

ADVECTION ALONE

• 1. Three dimensional advection test case (to be developed by Christiana Jablonowski)

TRANSPORT TESTS

• 1. Tracer tests involving the maintenance of ratios, etc. (to be developed by Phil Rasch)

BAROCLINIC CORE ALONE

• 1. Jablonowski baroclinic instability test case (NCAR/TN-469+STR)

Provide error norms for the steady state component

L2 p_s difference norms against reference solutions At what resolution does the core produce the reference solutions to within their uncertainties p_s amplitude and phase errors at lower resolutions

Williamson and Jablonowski can provide netcdf files of reference solutions or calculate the norms if provided with netcdf files of data on Gaussian grids or equally spaced latitude/longitude grids with pole points.

• 2. Giraldo and Rosmond Rossby-Haurwitz wavenumber 4 (MWR, 2004, pp.133-153) Based on Monaco and WIlliams, NPS TN, 1975 (volunteer needed to determine desired analysis)
• 3. Flow on the sphere with a isolated mountain from: Smolarkiewicz, Margolin and Wyzogrodzki, 2001: A Class of Nonhydrostatic Global Models, J. Atmos. Sci., 58, 349-364 (to be further developed by Christiana Jablonowski)
• 4. Atmospheric sudden warming case which sets up flow over the pole 17 February 1979 to 5 March 1979 25 February has strong flow over the pole (details to be specified by Dave Randall)

BAROCLINIC CORE WITH IDEALIZED FORCING

• 1. Held-Suarez (BAMS, 1994, 75, 1825-1830)

Overall behavior as with other AGCMs. i.e. "climate modelers' norm" applied to zonal averages

BAROCLINIC CORE WITH CAM3.0/3.1 PARAMETERIZATIONS

• 1. Aqua-Planet - APE "CONTROL" case (at proposed application resolutions) Neale and Hoskins, ASL, 2000, 101-107

A "true" solution is not known.

Look for behaviors that are not inconsistent with those of Eulerian, semi-Lagrangian and Finite Volume cores of CAM, and/or within the range of acceptable APE participants.

Some diagnostics of interest are zonal averages, KE spectra, wavenumber-frequency diagrams, precipitation pdf.

(Note: this is almost as complicated as the full model, but does avoid the complexity of coupling to the land and sea ice models of the stand-alone CAM. It is suggested that this run be done in the software engineering environment of the CAM to be sure the parameterizations are actually the same and to take advantage of the history file capabilities of CAM since the standard APE diagnostics are rather extensive. CAM3.0/3.1 parameterizations are chosen because there are many APE runs with different resolutions and different dynamical cores available for comparison. Results from these will be published in the future.)

ROUNDING DIFFERENCE GROWTH

• 1. Growth of rounding-level (10^-14) perturbations should be shown to be modest as expected for unforced baroclinic flow when starting from atmospheric-like states such as one from an APE run (above) or from a climate run with complete parameterization set. But model is run adiabatically, i.e. only growth from the dynamical core. (See short dashed line in Fig.2 of Rosinski and Williamson, 1997: The accumulation of rounding errors and port validation for global atmospheric models. Siam. J. Sci. Comput., 18, 552-564.)

---

GENERAL COMMENTS

It is generally useful to have an indication of computing resources required for the tests and an estimate of what they might be after additional software engineering, i.e. optimal code is not needed for the tests listed above at this stage.

As the tests suggested above are agreed to by the working group, benchmark runs can be made with the dynamical cores available in CAM at several resolutions. (Some do exist already)

Back to CAM Development Plan

Back to AMWG Homepage