Home » Technology » Scientific Computing » Back to the Bay

Back to the Bay

© Adam Hughes

As mentioned in last week's article, one of the main areas of computational focus at the San Diego Supercomputer Center is the discipline of Earth Systems Science. Serving as the testbed for many computational ecological techniques, the San Diego Bay Modeling project is the spearhead of the efforts in this endeavor. This project employs three main levels of computational methodologies : 2-D Hydrodynamics, 3-D Hydrodynamics, and a model of watershed runoff. Last time, we examined the basic tenets of this modeling project and the 2-D Hydrodynamics tools it employs. This time around, we'll take a look at the 3-D Hydrodynamic method used in this endeavor.

The three-dimensional hydrodynamic model used in this study is called ECOMsi. ECOMsi is a regime for modling both coastal and estuarine circulation that was developed by Blumberg and Mellor. The basic hydrodynamic differential equations are solved via finite difference methods to yield such properties as free-surface elevation, water temperature, salinity, and quantities related to water turbulence.

As might be imagined, ECOMsi shares many characteristics with the 2D hydrodynamic model discussed last week (TRIM2D). In particular, the grid used for simulation purposes is similar to the TRIM2D grid. It measures 20.0 km by 15.6 km with a horizontal resolution of 100m. The third dimension is constructed by placing multiples of the above plane on top of each other, usually about ten deep. The exact parameters used in these simulations are derived from data provided by the Navy, the Army Corps of Engineers, and NOAA.

Typical simulation times here are again 3 to 30 days, with a time step of 20 seconds. This time step is significantly smaller than the 4 to 6 minutes used in the TRIM2D simulations, a fact which is not too surprising when the added dimension is considered. Generally, the more complicated a model system is, the more difficult it is to get it to behave in some physically sensical way during simulation. One way to bring control back to a time-dependent study is to reduce the time step so that the system doesn't undergo wild instantaneous fluctuations which tend to send its overall energy into the stratosphere or into a black hole.

With increasing system complexity, equilibration can also become more of an issue. Because model systems are often set up as an input file from one set of initial coordinates, they usually represent one thin slice of the lifetime of the real system. As such, there can be gross inaccuracies of constituent spatial relationships, such as particle superpositioning, even in a configuration that is completely viable and will eventually run fine. The standard procedure

The copyright of the article Back to the Bay in Scientific Computing is owned by Adam Hughes. Permission to republish Back to the Bay in print or online must be granted by the author in writing.