Modeling the arteries of the land

You don’t always need sophisticated computer models to simulate the natural world. Take the routing application for parallel computation of discharge (RAPID) model, for example. A relatively simple mathematical equation within RAPID is used to simulate river flows moving downstream in massive river networks, such as in Mississippi, US, and much of Western Europe. 

These models are used to compare computations with historical observations to account for how rivers are connected and also help improve understanding of severe droughts. “RAPID uses a simple equation to compute flow in rivers. Many other models use much more advanced equations but are usually applied to much smaller areas,” says Cédric David, a scientist who helped develop RAPID and recently started working at the University of California Center for Hydrologic Modeling.

One example of a complex water model is LizzaPAKP. But, this uses complex equations to focus on smaller areas of groundwater flow, as opposed to RAPID which looks at much larger systems. “We provide an estimate of flow over thousands of connected mapped rivers. I don't know of another model that does this,” says David.  

The latest research papers by David and his colleagues are published in the Journal of Hydrometeorology and in the Journal Hydrological Processes. Their model has been applied to areas of 100,000 km² (38,610 miles²) to 1,000,000 km² (386,101 miles²). So far, the biggest computing problem the researchers faced was modeling the Upper Mississippi River Basin, which spans about 2,092 kilometers (1,300 miles). Their simulation used 182,240 computing elements from resources at David’s geosciences department and the Lonestar supercomputer at the Texas Advanced Computing Center. 

The programming language used is the same one that meteorologists, or TV weather people, use to forecast the atmosphere – Fortran. RAPID enables easy incorporation of these complex atmosphere and land surface models, which are run together with its simple equation of downstream water movement to give a complete picture of water distribution over countries and continents. 

“In general, the ability to apply high-resolution solutions to river hydraulics is important (think flash floods), but I would hasten to add that computing is just one part of it,” says Richard Hooper, executive director of the Consortium of Universities for the Advancement of Hydrologic Science, Inc, who is not involved in this work.

Human factors can also play an important role in natural disasters. A recent SciDev.net article suggests that the lack of rapid human action from an impending flood, for example, may be caused by the technical language used by warning systems, with individuals and communities perhaps not able to understand the meaning of obscure terms. It is suggested an integrated approach of technology, communication, social sciences, and psychology will address this issue.

David says he hopes his research will make a real impact on policy. “We've been actively collaborating with the Texas Commission for Environmental Quality and even the French Weather Service, MeteoFrance, to see how such a model could be used operationally by decision makers.”  

Now, David and his colleagues hope to test a near real-time application of RAPID by the end of 2012 and are very keen on forecast applications, but first they have to figure out the best way to leverage parallel computing resources by more efficiently splitting computations between cores.

- Adrian Giordani