ROMS as a Component of the
Community Climate System
Model (CCSM)
Enrique Curchitser, IMCS/Rutgers
Kate Hedstrom, ARSC/UAF
Bill Large, Mariana Vertenstein,
Don Stark, and Jon Wolfe, NCAR
October, 2007
Climate Model Biases
The Regional Setup
SST, August 18, 2000
Need to downscale atmosphere as well!
9+27 km forcing COAMPS
~ 200 km forcing (CORE)
NEP Implementation
• 10 km average horizontal resolution
• Run10:
– 42 vertical layers
– Coupled sea ice and ocean
– Six-hourly fluxes from Common Ocean-ice
Reference Experiments (CORE) reanalysis
– IC’s and BC’s from CCSM (POP) forced by CORE
– 1958-2004
Evaluation of Latest
Simulation
• We are much happier with the heat
fluxes from CORE than from NCEP
• Ice area seems quite realistic
• Still could do better:
– Not enough fresh water input at the coast
– Stratification in general isn’t quite right, will perhaps
be improved by tidal mixing, different IC/BC
– We want a shallower minimum depth
• We also need to add an ecosystem
model for GLOBEC
Ice Area and Volume
CCSM Coupling
• Thanks to the CORE group’s products,
we no longer need to run the large NPAC
domain (proof of one-way coupling)
• The CCSM model does fine for its
resolution, but fails in regions needing
more resolution
• The idea is to couple ROMS to CCSM,
where ROMS is part of a “composite
ocean” providing a better sea surface
temperature to the atmosphere
CCSM
Standalone ROMS (ocean.h)
#include “cppdefs.h”
PROGRAM ocean
USE …
#ifdef MPI
CALL mpi_init
CALL mpi_comm_rank(…)
#endif
CALL initialize
CALL run
CALL finalize
#ifdef MPI
CALL mpi_finalize
#endif
END PROGRAM ocean
Standalone ROMS (nl_ocean.h)
• Included by ocean_control.F
• Contains initialize, run, finalize for
the nonlinear ocean model
• Run calls main3d or main2d inside
the timestepping loop
• Calls get_data which reads files
(both in initialize and in run)
Coupled Mode
• Composite ocean knows input
filename, NtileI, NtileJ, dt, and
ntimes
• Two phase ROMS initialization,
first calls inp_par, second calls
get_data
• get_data copies fields from CAM
• run is called each coupling
interval
CCSM Strategy
• The coupling interval is one day, with
the atmosphere running ahead
• The composite ocean will receive
hourly winds, etc. and interpolate to
ROMS grid(s)
• POP runs for that day, sends “curtains”
out for ROMS nests
• ROMS runs for that day
• Composite ocean merges SST from POP
and ROMS for CAM
Concurrent Execution
Ice
From Dan Schaffer
Sequential Execution
Ice
From Dan Schaffer
• CCSM is now concurrent, with
differing numbers of processors
for each of the components
• An effort is underway to make it
sequential, for the IBM Blue Gene
and other such computers
• The ROMS coupling strategy
described is more sequential
Patches
• More than one simultaneous patch
• Can now handle through Ngrids
feature
• Trouble:
– Southern ocean (ICE and ICE_SHELVES)
– Bering (ICE, TIDES, and ECOSYSTEM 1)
– Peru (ECOSYSTEM 2)
• What to do?
Conclusions
• We are excited about this opportunity
to work with the CCSM group
• A similar effort is underway to add
WRF as a regional atmospheric model
• Very much a work in progress, still to
be done:
– POP “curtains” to ROMS
– SST back to composite ocean
– Changing to CCSM bulk_flux algorithm