Update from Will (Brian, Christiane, all)

  • open software issues raised by Will (from web page)

Will be discussed off-line; Brian in particular needs
time to read through the issues.

  • spun-up initial data file for ideal_phys runs

The spun up runs (first 7.25 days) for ideal physics
appears good.

  • ideal_phys output issue?

Solved by Christiane and Brian. Use default
history file names (i.e. do not define hfilename_spec)
or set namelist variable MFILT to a sufficiently large
value.

  • C46 versus C48: code crash in C48

No specific ideas on this problem. Will will try to
pinpoint where it is crashing and why.

  • status of adiabatic test case implementation:
    are all test cases including the 3D advection implemented via namelist switches?

No! We only have a selection of the following:

! 10 = hydrostatically balanced 3D test with idealized mountain
! 11 = Use this for cold starting the climate model with USGS terrain
! 12 = Jablonowski & Williamson Baroclinic test case (Steady State)
! 13 = Jablonowski & Williamson Baroclinic test case Perturbation
! 14 = Use this for cold starting the Aqua-planet model
! 15 = Small Earth convective bubble
! 16 = 3D hydrostatic non-rotating Gravity waves
! 17 = 3D hydrostatic rotating Inertial Gravity waves (case 6-3-0)

Of these, only cases 12 and 13 appear to work – in other
cases it seems that some of the data are not properly defined
(crash at initialization). Bill only guaranteed that the
baroclinic tests would work. Elaboration?

See discussion below in "Plan for the next few weeks".

  • status of p_d_coupling for non-adiabatic runs, interpolation of tendencies (ideal_phys, aqua-planet)

There are problems with aqua-planet (see discussion in email
"What works and what doesn't". But Ideal Physics seems to be
working (see results), implying that p_d_coupling is not the
problem.

  • are there open issues with use of pio (Jim, Will)

PIO itself seems to work as advertized. There is a bug
in writing or reading the restart file (see separate discussion)

  • documentation of configure and namelist settings, machine-dependent?

Not discussed due to connection problems. Important – needs
to be discussed at next meeting.

  • are there machine-dependent compilation and runtime issues?

Yes. Needs to be discussed at next meeting.

  • status of FV cubed-sphere restarts

All modes write restarts without any apparent problems.
In restart runs, however, the read_restart_dynamics routine
appears to fill the state correctly, but the code is
dying in the call to fv_dynamics (the GFDL fvcubed dycore
interface). Investigations ongoing – it is possible
the data are read back in some permuted fashion.

  • Will's visit to UM and NCAR

Dates: UMICH (May 19 - 22), NCAR (May 25 - June 1)

Post-processing of FV cubed-sphere data (Will, Mark)

  • interpolation to lat-lon: HOMME's post-processing routines or NCL?
    • HOMME stand-alone interpolations to lat-lon grid (not applicable to FV^3.
    • NCL's internal interpolation, using NCL scripts (similarly used in FV^3)
  • Mark's NCL scripts for aqua-planet runs

Use of NCAR wiki:

Computing aspects, next steps (Art, Pat)

  • Current performance

Was this discussed after Will's phone-line went down?

  • Additional OpenMP parallelization?

Was this discussed after Will's phone-line went down?

Plan for the next few weeks:

  • run full suite of the adiabatic test cases

GFDL internal test suite is not sufficient. This is
high priority for the CCSM workshop. Will is going to
patch in J&W test_standalone for a full suite.

  • is the 3D advection test (slotted ellipse) implemented?

No! See solution in previous point.

  • long-term 'ideal_phys' Held-Suarez simulations

Will be run immediately now that history file problem is understood.

Participants should check WIKI documentation for any omissions.

  • plans for the CCSM meeting in June, poster

Suggestion would be to have Christiane present a talk
on the progress, as long as adiabatic and ideal physics
runs are giving reasonable results (should know this by
next week).

  • No labels