Foreseeing The Deluge

The forces of nature are not controllable, and thus natural disasters are not preventable. However, improving our ability to predict natural disasters can drastically reduce the associated loss of human life. Increasingly, advances in technology are making disaster forecasting more available and sophisticated.

The Cévennes storms
During the last decade, flash floods have become one of the most significant natural hazards in Europe. One example occurred in 2002, in the Cévennes mountain range in south central France. During the autumn, the Cévennes mountains draw warm, humid lightning storms in from the Mediterranean. Depending on the cloud system, the storms can remain in the area, stuck churning in the mountains, dumping moisture, for many hours. In 2002, one such storm lasted 15 hours and caused 600 mm of rain equivalent to a year’s rain in just one day. Resulting flash floods caused the deaths of more than 20 people and economic damage estimated at 1.2 billion Euros. Five hours later, the storm water had cleared from the land. “There were no warning systems set up in 2002,” says Vincent Thierion, a geo-informatics researcher working on a grid-enabled flood forecasting application. “Such events had been rare, and now they are more common, occurring also in 2005, non-catastrophically, and last fall in 2008. You have no chance of knowing that one could happen until about 10 hours before.”

Advanced warnings save property and lives
Thierion works on the European project CYCLOPS—Cyber-Infrastructure for CiviL protection Operative ProcedureS—a project aimed at bridging the gap between European civil protection agencies and the grid computing community. Supporting the work of the French Grand Delta flood forecasting service, CYCLOPS has developed a platform for running the forecasting application G-ALTHAIR on a distributed computing grid and ported this to the Enabling Grids for E-sciencE infrastructure. This style of computing allows the application to multi-task: researchers can simultaneously run forecasts for several watersheds, and for several rainfall scenarios, as well as incorporate new data or add additional layers of forecasting complexity in real-time. CYCLOPS researchers hope this will aid authorities in making decisions when managing crises. Thierion believes G-ALTHAIR, currently in prototype phase, will be in production by 2010. “We met with the civil protection service a few weeks ago and they are very interested in this prototype, so we must work to finish it so they can use it!” The application uses data from about 170 regional monitoring stations spread around France’s floodprone zone, as well as radar images and rainfall forecasts from MeteoFrance, using grid computing to simulate the real-time hydrological situation, and to forecast possible scenarios. “The existing system is mostly dedicated to monitoring a flash flood which is already occurring,” Thierion says. “With G-ALTHAIR, the most important innovation is use of the grid to prevent damage. We can share richer hydrological information with the civil protection services, to protect the population.” Thierion demonstrated G-ALTHAIR at the 4th Enabling Grids for E-sciencE User Forum in Italy in early March. Watch the demonstration through the GridCast video blog (http://gridtalk-project.blogspot.com/2009/03/cyclops-hydrology-app-fights-flash.html ) or learn more about CYCLOPS at http://www.cyclops-project.eu/