Custom Query (964 matches)

There is distributed-memory bug in routine ana_nudgcoef.h when setting the passive/active coefficient used in the nudging terms for open boundary conditions. We need to use mp_collect instead mp_bcastf. This is a global reduction opearation instead a broadcast communication. See the following ​forum post for more details.

Many thanks to Francesco Nencioli for reporting this problem.

The W4DPSAS weak-constraint data assimilation algorithm was redesigned to follow closely Courtier's (1997) dual formulation paper. This upgrade includes:

• Added spectral and Ritz preconditioning to W4DPSAS similar to that coded for W4DVAR.
• Added hot start for subsequent outer loop in W4DPSAS and W4DVAR. This can be activated by turning new input parameter LhotStart in s4dvar.in.
• Added all the variables necessary for the observation sensitivity driver in output NetCDF MODname. We are currently working on this new driver which is the adjoint of the W4DPSAS and W4DVAR system.
• Removed obsolete routine impulse.F.

Now both weak-constraint data assimilation driver are working very nicely. Many thanks to Andy Moore for his help coding and testing these algorithms.

The local variable Iinp was not assigned in routine wrt_impulse.F.

Also corrected a problem in build.bash when the macro definitions are commented out. We need to add a colon when testing if the macro expands to NULL:

We need to use:

if [ -n "${USE_MPIF90:+1}" ]; then
  ...
fi

indead of

if [ -n "${USE_MPIF90+1}" ]; then
  ...
fi

Many thanks to David Robertson for fixing this problem.