Section title: From parent to child
go to page 1
go to page 2
you are on page 3
go to page 4
Title" West Coast Models, by Karen Green

For some ROMS models, the researchers use field data to set the initial conditions; for others they create hypothetical conditions. The simulation is then built as a series of grids of different resolution used to measure variables that must be examined on different scales. These nested grids work much like Adaptive Mesh Refinement techniques. The simulation starts with the coarsest resolution grid (the parent grid), which measures a mesoscale phenomena. Finer resolution "child" grids are then embedded within the parent.

 
click to enlarge
Instantaneous, summertime chlorophyll densities (measured in milligrams per cubic meter) showing coastal upwelling and mesoscale filamentation in the California Current System. The ROMS simulation uses a U.S. West Coast parent domain and a central California embedded child domain with horizontal grid resolutions of 15 and 5 kilometers, respectively. Their boundary is marked by the black line.

This hierarchical embedding of grids works on many resolution scales over many levels of embedding, explains McWilliams. For studies of the Pacific basin, for example, the researchers start with a very coarse-grain grid that measures at 50-kilometer spatial intervals and works down to a much finer local scale. For local studies around Monterey Bay and the Southern California Bight, the researchers use a three-level embedded grid. The first grid measures at 15-kilometer intervals. Advancing this parent grid by one step determines the boundary conditions for the second level of the grid--in this case a grid with a 5-kilometer resolution. A third grid (the second child grid) looks at phenomena at a 1.5-kilometer resolution. The simulation continues by advancing each child grid forward in space and time enough so that it covers the same space and time as the parent grid. The parent grid is then updated with the information obtained from the child grids to create a more accurate model that measures both large- and small-scale events over time.

For example, to understand how a major winter storm affects the movement of sediment in a local area of the coastal system, a ROMS researcher might look at the regional currents on one level, outflow of river water and debris into the system at a second, finer level, and localized eddies at a third, still finer level. Using an embedded grid, says McWilliams, researchers can see the impact each phenomenon has on the movement of sediment as well as how the phenomena interact and influence each other.

So far, the ROMS research team has run simulations that use as many as four grid levels, with the finest grid measuring at 500-meter intervals in the Santa Monica Bay. The embedded method can handle an unlimited number of grids, but the finer the grid, the more computationally intensive the simulation becomes, explains Patrick Marchesiello, a member of the ROMS research team. For example, it takes 27 timesteps at 1.5-kilometer resolution to equal one timestep at 50-kilometer resolution.

"Embedded gridding can be done on many levels. A very large-scale event like El Niño would be measured on a scale of 100 kilometers," McWilliams says. Although ROMS is a regional modeling system, a similar global ocean modeling system could be developed when the computing power to support it exists.