Graduate Student Receives Prestigious National Weather Service Fellowship

October 22, 2007

AUSTIN, Texas—The National Weather Service (NWS) has awarded Enrique Rosero, a PhD student at the University of Texas at Austin’s Jackson School of Geosciences, a graduate fellowship worth $50,000 a year to support his work to improve computer models that the weather service uses to forecast flooding.

More Americans die each year from floods than any other severe weather related hazard. So the NWS continues to look for ways to improve the accuracy and lead time on flood warnings across the U.S. Rosero says many of the concepts and technologies he will use have already been developed.

“What makes this unique is the integration of different components that already exist independently into a coherent framework that enables us to have a system that works as a unit,” he says.

The NWS currently uses the River Forecasting System (NWSRFS), an integrated set of computer programs first developed in the 1970s (with some modifications through the years) that use near real time precipitation data to generate probabilistic forecasts for stream flow.

More Americans die each year from floods than any other severe weather related hazard. Image: NOAA/NWS

The system has its weaknesses, though, such as handling high-flow events or spatially distributed mixes of rain and snow. The system makes many compromises in simulating the features of the real world, sometimes because the underlying physics are not well understood and sometimes to reduce computational needs. For example, impervious land cover—land surface that is unable to absorb water—is not accurately distributed throughout a region, but instead taken as a lump sum across the whole region.

“So you’re just taking a chunk of available land out,” he says. “But it makes a difference where you take the chunk out, whether it is in the upstream part of the catchment or the downstream part, because it’s going to slow down or speed up the velocity of water waves. With the new models, we can take into account this heterogeneity.”

To more accurately model the real world, Rosero will add to the existing system two models coupled together: a high-resolution, distributed hydrologic land-surface model developed by the National Center for Atmospheric Research called Noah-d-LSM and a dynamic, flood-wave routing model developed by the Department of Defense called HEC-RAS.

“This was a dream since the ‘40s, to have a model that represents all the components and that works in the small pieces and gives you the entire answer,” says Rosero. “With a better understanding of the underlying processes and advances in computational muscle, that is now possible.”

A second enhancement involves new techniques for calibrating the models that adjust for errors in observational data and tweak fundamental model parameters under uncertainty.

A third enhancement will take three core models in the system (the original model at the heart of the NWSRFS, plus two added by Rosero) and produce a probabilistically weighted average. Models that do a better job of mimicking the real world over time will be given more weight than others in future runs. This is known as Bayesian model averaging.

Rosero, who competed with students from across the U.S., received one of only two NWS Office of Hydrologic Development Graduate Fellowships. The award is all the more prestigious given that the other fellowship was reserved for an internal candidate.

Rosero is in the right place to carry out this work. The University of Texas at Austin is home to Lonestar, one of the world’s most powerful supercomputers. Next year, the university’s Texas Advanced Computing Center will bring online an even more powerful supercomputer named Ranger. Since arriving at the university in January, Rosero has already run simulations on Lonestar.

“In my application, I was very specific in saying this is very expensive, but we can do it here,” he says. “I’ve done it before and this is what I expect. With the implementation of a new generation of software and computer at UT, we can afford it.”

His application was also bolstered by his collaboration with advisor Zong-Liang Yang, associate professor and head of the Land Environment and Atmospheric Dynamics (LEAD) group, one of the Jackson School’s primary climate research and education groups. Yang has worked at the forefront of climate modeling for many years.

Rosero predicts the task of integrating the various models and techniques will take about two years. He says the first real test of the new system will be made using actual observations from the Guadalupe River Basin in south central Texas over the past year. The heavy rains and extensive flooding that occurred in many parts of Texas could dramatically highlight the value of such a system.

“I think it will raise awareness in decision makers if we can show the governor, ‘Look, our model could have predicted this,’” he says. “I think they will want to invest in our research.”

By Marc Airhart

For more information about the Jackson School, contact J.B. Bird at jbird@jsg.utexas.edu, 512-232-9623.