Crisis Management in Bays and Estuaries

PARTICIPANTS
Clint Dawson, Victor Parr, Robert Kirby, Jennifer Proft, Sharon Lozano, CSM/TICAM, University of Texas at Austin

n oil tanker slips through the Houston Ship Channel at the mouth of Galveston Bay on its way north to the port of Houston, when suddenly something goes wrong. Perhaps the tanker hits a submerged obstacle, or an explosion rocks the huge vessel. In any case, something breaches the hull and a huge oil spill begins. Where will the incoming tide take it? How and how fast can all concerned contain the spreading environmental disaster? A new NPACI Earth Systems Science project called "Crisis Management in Bays and Estuaries" is seeking the answers to these critical questions.

IMPORTANCE OF BAYS AND ESTUARIES

FLOW AND TRANSPORT CODES

JOINING SIMULATIONS

IMPORTANCE OF BAYS AND ESTUARIES

The project unites a team of geophysical researchers from the Center for Subsurface Modeling (CSM) at the University of Texas at Austin, led by CSM director Mary F. Wheeler, with a group of computer scientists at the University of Maryland, led by Joel Saltz and Alan Sussman--participants from NPACI's Programming Tools and Environments thrust area.

The natural harbors of the United States include more than 30 major bay-and-estuary systems. The coastal strips and shallow waters of bays and estuaries are unique ecological niches for multitudes of species. Shorebirds, sea mammals, fish, mollusks, and crustaceans all find in these zones their main habitats and means of sustenance. But busy ports, with their urban sprawl, toxic outflows, and oil spills, can pose huge threats.

Galveston Bay is an excellent example of such a system. It sits at the confluence of many winding bayous and river systems, with the port of Houston at its northern end and Galveston on the westernmost of three islands at the mouth, which surround a narrow channel that is the major inlet for shipping. The shallow waters are dotted with commercial oyster beds. A dozen local, state, and federal agencies are now cooperating with university-based scientists to develop means to evaluate the health of the bay and its vulnerability to crises like the oil spill described above.

"What we are building," said Wheeler, who was elected to the National Academy of Engineering in 1998, "is timely, efficient, and accurate simulation and data management for modeling such crisis situations as environmental catastrophes in bays and estuaries."

Figure 1. Oil Spill in the Houston Ship Channel

For a hypothetical oil spill, a water flow simulation has been linked to a projection code and transport simulation to show how an initial oil spill (top, in red) in the Houston Ship Channel spreads on an incoming tide (bottom). The simulations are from a group led by Mary Wheeler at the Texas Institute for Computational and Applied Mathematics (TICAM), while the connections between them are made possible by the Active Data Repository project, led by Joel Saltz at the University of Maryland. The linked simulations produce data that can be mapped and visualized as the calculations proceed, as these two frames illustrate.

FLOW AND TRANSPORT CODES

The task requires the kind of collaboration that NPACI was designed to foster--coordinated effort by both computational and computer scientists. Wheeler's CSM is part of the Texas Institute for Computational and Applied Mathematics (TICAM). The close-knit team of faculty and research scientists has expertise in applied mathematics, engineering, computer science, and physical, chemical, and geological sciences. Their models demonstrate how high-performance parallel processing can be used as a tool to discover the behavior of fluids in permeable geologic formations--such as petroleum and natural gas reservoirs, groundwater aquifers, and aquitards--and in shallow water such as bays and estuaries.

The group has developed two simulators, UTBEST and PADCIRC, that use various finite-element schemes to capture the circulation patterns in shallow coastal seas, bays, and estuaries. PADCIRC (for Parallel Advanced Circulation Model), an efficient, parallel simulator for distributed memory platforms (SPMD machines), has been built upon ADCIRC (Advanced Circulation Model) created by professors Johannes Westerink of Notre Dame and Rick Luettich at the University of North Carolina. Its grid for Galveston Bay, based on unstructured triangles contains 2,113 nodes and 3,397 elements.

"We've been keeping track of PADCIRC's scalability," said Clint Dawson, associate professor at TICAM. "If the problem can be made large enough, it scales very well to 256 processors." The codes were developed by Wheeler, Dawson, and collaborator Srinivas Chippada of Fluent Technologies, working with programming consultant Victor Parr and graduate students Robert Kirby, Jennifer Proft, and Sharon Lozano.

For a complete simulation of Galveston Bay, the hydrodynamics simulator needs to be coupled to a chemical transport simulator that uses the hydrodynamic input to simulate the time-dependent transport of chemicals--such as an oil spill--within the domain. Since the chemical reactions have little effect on the circulation patterns, the fluid velocity data can be generated once and used for many separate simulations.

Here, too, the Wheeler group has used two models. One, CE-QUAL-ICM, was developed by Carl F. Cerco and Thomas Cole of the Army Corps of Engineers Waterways Experiment Station in Vicksburg, Mississippi. Modeling more than 20 state variables, including sediment transport, it is based on an unstructured finite-volume method, explicit in time, and can be run in one, two, or three dimensions. The Wheeler group has developed a parallel version of CE-QUAL-ICM called PARWQM. The other, UT-TRANS, developed in the Wheeler group, computes advection, diffusion, and reactions of contaminants in shallow waters, using a mass-conservative, upwind finite-volume method and explicit time-stepping.

JOINING SIMULATIONS

For many reasons, coupling any two of these simulators to form a complete system is not a straightforward process. The chemical transport codes have been developed for different grids and time dependencies. Thus the hydrodynamic information they require from the circulation codes must be preprocessed and projected onto the new transport grids.

Here is where the Maryland team was able to supply the most flexible and useful answers. Saltz, leader of the Programming Tools and Environments thrust area, assistant research scientist Alan Sussman, postdoctoral researcher Tahsin Kurc, and graduate student Chialin Chang have developed an Active Data Repository (ADR) to integrate the storage, retrieval, and processing of multiple multidimensional data sets on parallel machines.

The heart of the system is the code T2, which can simultaneously manage and process multiple data sets, like those from the Texas codes, despite their different attribute spaces and different distributions of data within each attribute space. The architecture includes an attribute space service, data loading, indexing services, data aggregation, a query interface, query planning, load balancing and memory management, and query execution. To all this, the applications scientists add appropriate translators from code to code.

Now it has all been put together and subjected to its first tests. In a typical run for a hypothetical oil spill in the Houston Ship Channel, PADCIRC generates water flow output (velocities at the grid nodes), which is stored in the T2 database. When CE-QUAL-ICM or UT-TRANS requests hydrodynamic output from T2 to compute the transport of the oil spill on an incoming tide, T2 retrieves and aggregates the water flow data, then sends the data to UT-PROJ, the projection code developed to work with the ADR by the Wheeler group. The output of the UT-PROJ is then sent to CE-QUAL-ICM, which performs the transport simulation, producing data that can be mapped and visualized as the calculations proceed. The results show how an initial oil spill in the Houston Ship Channel spreads on an incoming tide (Figure 1).

Because T2, the hydrodynamics code, and the transport codes are all parallel programs, they can use Maryland's Meta-Chaos runtime library to move large amounts of data between the programs (see page 12). Because the ADR performs such a large variety of pre- and post-processing tasks, it is possible to demonstrate the entire system performing in real time over the Internet despite the size of the codes involved, and such a demonstration was run at SC98 in Orlando and at the January 1999 NPACI All-hands Meeting.

"When these models are sufficiently refined, we will be able to respond to an actual spill," Wheeler said. "Given its location and the current tidal conditions, we can accurately predict how it might spread and where booms might be laid to contain it. Getting there requires tremendous cooperation of the sort fostered by NPACI, between applications programmers like ourselves and infrastructure developers such as those in the Maryland group." --MM