iSGTW Feature - Flood of data helps prevent hurricane damage

Feature - Flood of data can help prevent hurricane damage

These scientific visualizations of Hurricane Katrina were created at the LSU Laboratory of Creative Arts + Technology (LCAT) by the CCT sci-viz group.

New Orleans Perspective from Lake Pontchartrain, LIDAR elevation, GOES-12 satellite imagery, and Adcirc sea elevation and levee system, AUG 26 - AUG 31st.

The LIDAR heightfield is color coded: yellow/green for land above sea level, blue at sea level and violet below sea level. The land above sea level in New Orleans was formed by the Mississipippi River naturally flooding and depositing sediment. The natural levee that surrounds the river can be seen in green as well as the Gentilly, Metairie Ridge. The height of the storm surge is indicated by dark blue. The Adcirc levee system is shown in pink.

Image courtesy of

When the National Hurricane Center issues a hurricane advisory, emergency teams have little time to predict the locations and effects of storm surges and waves in order to identify areas that should evacuate. A prototype system, known as the Southeastern Universities Research Association Coastal Ocean Observing and Prediction (SCOOP) program, promises to run hurricane computer simulations and produce results quickly.

The complex hurricane models crunch very large data sets, and the data must move from storage to compute resources and back as quickly and reliably as possible. The SCOOP program relies on the Stork Data Scheduler software package - so called because it delivers data - to manage data placement and movement. Developed by researchers at Louisiana State University and the University of Wisconsin-Madison, and made freely available for download, Stork allows researchers to efficiently store, access, share and transfer large data sets.

"There are 80 or more model configurations that we currently run very quickly for our hurricane predictions, and efficient and reliable data movement is extremely important," said Gabrielle Allen, SCOOP collaborator and associate professor at the Louisiana State University Center for Computation and Technology and Department of Computer Science. "Stork automatically chooses the best parameters and transport mechanisms so the data is transferred in the most efficient way."

Distributed computing has historically focused on managing computing resources rather than data. Now that research has become more data intensive, inefficient data movement often creates a major bottleneck. Stork works with high-level batch schedulers to schedule computation and data movement tasks synchronously in one integrated system.

New Orleans Perspective, MM5 and Adcirc sea elevation and levee system, MM5 vectors show how the surge enters New Orleans from the drainage canals connected to Lake Pontchartrain and through the Intercoastal Waterway connected to Lake Borgne and the Mississippi River Gulf Outlet canal. Dark blue represents the height of the storm surge which reached a height of 15 ft in the Intercoastal Waterway and 8 ft in the drainage canals.

Image courtesy of

"Batch schedulers, such as Condor, specialize in scheduling computational tasks, but do not specifically consider data scheduling or data movement," said Tevfik Kosar, Stork project leader and LSU computer science assistant professor. "Stork allows users to schedule and optimize data transfer tasks within a basic batch scheduler environment."

Stork automatically verifies that files transfer correctly, and its error recovery mechanism ensures completion of the data transfer even if the initial transfer fails. Stork developers plan to release an update this spring with several new features, including caching of multiple data transfer requests and up-front estimation of completion time for a requested data transfer.

"Data-driven science applications, like the SCOOP hurricane scenario, illustrate the critical need for new tools that fundamentally support data management," Allen said. "It is clear that integrating data-centric technologies, such as Stork, into existing distributed compute resources such as the Open Science Grid or the TeraGrid, will lead to a fundamental shift in how we undertake research in science and engineering fields."

-Amelia Williamson, for iSGTW