Version History

Additional details are found in later chapters.

2.3

Released August 2002.

• A simple formulation for atmospheric boundary layer winds is used to compute the surface fluxes.
• Vertical momentum advection terms are included in the momentum equations.
• Fourth-order horizontal diffusion is used in the momentum equations in lieu of the second-order diffusion. (See Section 4.6)
• New cloud parameterization is implemented, considering random overlapping of different cloud types. The resulting total cloud fraction is ensured to be less than 1. (See Section 4.3.1).
• The top of the model is raised to 150 HPa and the vertical velocity structure of the 1 baroclinic mode ( ) is held constant above the 280 HPa level for numerical reasons (See Section 4.1).
• The model domain is extended to S - N.
• Corrections to spherical geometry terms are included.
• A moisture dependence is added to the gross dry stability to simulate the cloud-top effect on . A cap is used so that the gross moist stability would not decrease to a lower value than the minimum value typical of the reference profile region.
• The interception loss Evapi is capped at of precipitation amount (as a numerical safety check)
• The surface pressure gradients can be diagnosed.
• Baroclinic velocity potential can be diagnosed.
• The artificial ocean boundary near the northern and southern boundary used in version 2.2 is removed and the sponge layers are not needed by default. A Fourier filter is used to damp small scale features at high latitudes.
• Available Reynolds SST has been updated from December 1998 to October 2001.
• All boundary data except the cloudiness are updated for the new grid dimensions. There are numerous gaps in the ISCCP cloudiness data at high latitude making an interpolation to the new domain difficult. Currently the cloudiness data are identical to the ones used by QTCM version 2.2. The missing high latitude are set to the zonal means of the closest existing latitude band at 58.75 N and 58.75 S, respectively.
• An option is added to use prescribed observed SST in masked regions and mixed layer ocean in non-masked regions (See Section 4.4).
• Scripts to run ensemble simulations are added (See Section 2.3.2).

Changes in code structure:

• The code is adapted to Fortran-90.
• All common blocks are replaced by modules.
• Example makefiles for different platforms are provided.
• With the simple atmospheric mixed layer the surface stress no longer relates to the resolved winds by a simple linear formula. Therefore the damping parameters (equations (4.33) (4.34) and (5.15) in [2]) have been replaced by the corresponding surface stress terms (equations (4.12) and (4.13) in [2]).
• Stress taux,tauy components are computed on u, v grids, respectively. Previously etc terms related to stress were used in the time marching on u, v grids while the arrays taux, tauy were diagnostic calculated for output on the T grid.
• The order of the indices of the cloud fraction variable cld was changed from ``lon, lat, cld-type'' to ``cld-type, lon, lat''.
• The output routines are rewritten. GrADS control files for the binary output files are automatically generated.
• Adding (removing) an output variable now only requires adding (removing) one subroutine call.
• The scripts to run the model are modified so that multiple runs can be made simultaneously. To this end each run is executed in a separate subdirectory of directory proc.
• The interval to write restart files can now be specified.
• Subroutines to diagnose surface pressure and its derivatives are added.
• When the model blows up, the instantaneous values of standard variables are output for diagnosis and the monthly means for the last month before blowing up are also written.
• The prognostic variable arrays are dimensioned as (nx,0:ny+1) to avoid out-of-bounds access of memory which led to errors on some computers (e.g., IBM-AIX).
• Added ntoutr and nooutr to the input parameters to specify time in days between restart files and days to skip before starting to write a restart file. Default write interval is one year.
• Surface winds us, vs from the atmospheric boundary layer model are added to the list of standard output variables.
• The restart file output names now include are changed to qtcm_YYYYMMDD.restart.
• The front-end solvers for the FATD package are in qtcm.f90 (used to be in fatdpkg.f). The old front-end solvers for Cartesian Coordinates are deleted. Users can refer to v2.2 or older versions for these or set .
• New preprocessor macros are

  DELTA_CO2

  Enable radiation code to use a non-s standard CO level.

  TOPO

  Include topography effects due to barotropic divergence.

  MXL_OCEAN

  Use mixed layer ocean.

  BLEND_SST

  Use readsst in mask regions and mixed layer ocean outside.

  NO_WADV

  Switch off vertical advection terms.

  NO_ABL

  Switch off atmospheric boundary layer.

  OBSCLD

  Use ISCCP observed cloud cover climatology.

  SPONGES

  Enable sponge regions.

  CHIWIND

  Diagnose velocity potential using irrotational winds.

  ENSEMB_INI

  Create initial condition files for ensemble runs.

2.2

Released November 2000.

Changes in radiation package clrad1.F:

• For solar radiation, we use maritime aerosol type with column optical depth 0.2. Compared to version 2.1, the downward solar radiation at the surface (FSWds) decreases and the upward solar reflection at TOA (FSWut) increases because of the reflection of aerosol.
• For longwave radiation, we implement a new set of coefficients to estimate effects of carbon dioxide. It is a linear relation and it is good for double CO experiments. For a case with triple and more CO , its accuracy will be reduced.
• We also add an effect from the forth cloud type AsAc+CuSc. A climatological value of cloudiness for the fourth cloud type is used and the cloud cover is absorbed into the coefficients for both longwave and solar radiation schemes.

Changes in code structure:

• The code now uses a fortran 90 compiler and its preprocessor. Preprocessor macros are used as model switches

  Possible macros are:

  CPLMEAN ..... Compute mean flux values for coupling with an ocean.
  LINEAR_T1C .... Use the linear calculation for reference moisture profile instead of the standard nonlinear calculation.
  YEAR360 ..... A year has 360 days instead of the standard 365 days (no leap years).
  FCF77 ....... Do not use F90 constructs such as array operations.
  REAL=real*8 Use this for global real*8 (double prec.) calculations.
  

• The default computation is now single precision. All internal double precision function calls are replace by their generic names. Define REAL=real*8 in the makefile to switch back to double precision.
• The default computation now uses a 365 day calendar. Set ntout (ntouti) to -30 to get monthly mean output.
• The nonlinear convective closure code has been merged into the standard qtcm.F code and is the default choice.
• To ease the output of the moist stable energy GMq is now an array defined in qtcm.h. and is one of the standard output variables for NetCDF output.
• If day0 is not specified in the input file driver.in the date saved in the restart file is now used to set the initial time. To this end the initial call of ``Timemanager'' in the driver is moved into ``qtcminit''.
• Major restructuring of the output routines.
• Fixed the Timemanager routine for the perpetual model. For SSTmode = ``perpetual'', Timemanager now sets ``dayofyear'' to day0 of the initial month (month0).
• Changed control of when to output mean values. i.e., loutpmean = mod(dayofmodel-noout,ntout).eq.0.
• All surface fluxes are now interpolated onto the model's temperature grid.
• The makefiles for different machines are merged. Useful compiler flags are provided in comment lines.
• The link of rdsst.F and mxlayer.F to ocean.F is removed and replaced by a new target qtcm_mxl for a mixed layer version.
• Added a NetCDF output routine. To use it one has to replace the standard makefile by makefile.netcdf. This output routine requires an f90 compiler and the netcdf library available from http://www.unidata.ucar.edu.
• The subroutines in driver.F have been split off into driver_util.F.
• The calendar routines in driver_util.F are rewritten.
• The coefficients for vertical advection of momentum (Vwijk) are added in qtcmpar.in; however, they are not used in the momentum calculations in this version because the corresponding climatology needs improvements.
• An artificial ocean boundary is imposed at the two outmost grid points near the north and south boundaries to reduce chances of model ``blow-up'' associated with numerical instability at the boundaries. Therefore, no sponges are needed near the boundaries.

2.1

Released May 1999. This is the version used in [2].

Major changes and additions from version 2.0:

• Vertical profile of b1 (the first basis function of moisture perturbation from reference state) is changed to avoid negative moisture in the upper levels (see [2] p17). Corresponding radiative coefficients in clrad1.f are updated.
• The meridional boundary of the model is stretchable up to the poles. All grids are shifted a half-grid point in the meridional direction. See Section 2.1.3 for details.
• A slab mixed-layer ocean model is implemented. See Section 4.4 for details.
• Two additional convective closures are implemented in qtcm.f.mconvct2. See Section 4.5 for details.

Minor changes from version 2.0:

• utilities.f is added. See Section 2.2 for details.
• bndry.f is modified to read multiple monthly boundary data and also to conduct time interpolation into daily data.
• FirstTime.sh is renamed spinup.sh.
• Several makefiles for different machines (SUN,CRAY and SGI) are provided. A set of shell scripts for running QTCM on CRAY is also provided.

2.0

Released August 1998. Major changes and additions from version 1.0:

• Method for calculating wind speed used in the evaporation parameterization is altered.
• Dynamical equations now account for spherical geometry. See Section 3.3.3 for more details.
• noout run-time option added.

1.0

Released July 1997. This version is the -version of the model. At the time of release, this version was numbered 2.0. In Summer 1998, it was retroactively renumbered 1.0; in the source code for v2.0 and later versions, this version is referred to as v1b. This was the version used in the Lin et al. [6] paper on the Madden-Julian Oscillation.

original

January 1997. Not publicly released. In the source code for v2.0 and later versions, this version is referred to as v1a. This is also referred to as the -version of the model.