The CSU Flux Coupler was designed to allow a general coupling of the climate subsystems in the CSU GCM. These are the atmosphere, the ocean and seaice, the vegetated land surface, and the permanent landice.  The purpose of the flux coupler was to allow each climate subsystem to be simulated at spatial and temporal scales appropriate to that subsystem, independent of the other subsystems.

The following philosophy guided the design of the flux coupler. The grid for the subsystem models should be independent of the atmospheric grid and each other. The time increments for the subsystem models should be independent of the atmospheric timestep and each other. Interpolation of fluxes of energy and water at the surface must be conservative and all interpolation should be accurate. The parameterization of physical processes that are closely coupled to the surface should be done at the resolution of the subsurface models. Of course, trade-offs with computational expense and complexity will cause some of the principles to be compromised to some degree.

Grid definition -
In the CSU flux coupler a global surface base grid is defined. This grid is logically rectangular, and is not restricted to latitude-longitude grids only, but can also represent stretched grids and offset poles (e.g. the POP ocean model grid). The resolution of the most highly resolved surface component is the resolution of this surface base grid. Surface base grid cells belong uniquely to one subsystem model and coastlines are defined at surface grid resolution. The subsystem model grids can be arbitrary, but the grid cell boundaries must coincide with the boundaries of the surface grid. Thus, a subsystem model grid cell can be considered an aggregate of the subsystem model grid cells and is the union of the subsystem model grids. Further, the surface grid cells within a subsystem model can be aggregated to form the actual subsystem model grid.

Interpolation -
A number of state and flux variables must be communicated between the atmosphere grid and the surface grid. Since the surface grid is of higher resolution variables are interpolated going from the atmosphere to the surface grid and averaged in the reverse direction. Two types of interpolation are available in the flux coupler. Bilinear interpolation is continuous and second order accurate everywhere, but does not conserve the global integral between the atmosphere and surface grids. It is suitable for interpolation of state variables such as surface pressure, temperature and water vapor mixing ratio. For conservative interpolation of fluxes, there is an interpolation package from the LANL ocean modeling group, which can do both first order and second order accurate interpolations. The first order interpolation results in step functions on the surface grid and so it is generally not used. The second order interpolation can extend the range of extrema. An adjustment is made to insure that certain limits are not exceeded for minima (e.g. maintaining precipitation or solar radiation positive definite). The averaging from the surface grid to the atmosphere is done by weighting the flux from the surface grid cell by the area in common with the target atmosphere cell.

Physical parameterizations -
The choice of physical parameterizations to compute on the surface grid in the CSU flux coupler is tied to the formulation of the planetary boundary layer (PBL) in the CSU GCM. The PBL in the CSU GCM is assumed to be a mixed layer. Mixed layer theory assumes fast communication of surface properties throughout the PBL through turbulent transport. Thus, in addition to the calculation of the sensible and latent heat fluxes and the surface drag coefficient on the surface grid, the flux coupler also computes the PBL parameterization on this grid. This includes the calculation of turbulent kinetic energy, the PBL top entrainment rate, the PBL top stratus cloudiness, and the exchange of energy and moisture across the PBL top due to Cloud Top Entrainment Instability (CTEI).

Temporal Coordination -
The temporal coupling is synchronous, and the most recent values of state variables and fluxes are communicated to the subsystem models every time step.

Donald Dazlich      
dazlich@atmos.colostate.edu