- August 26, 2007
- Posted by: EARSC
- Categories: Archive, European EO Industry
Monsoon season hit South Asia hard this year, with weeks of persistent rain causing rivers to break their banks and inundate the thatch-roof homes of India’s poorest provinces. Flooding and its wake — water-borne illnesses, hunger, snakebites — have killed 2,800 people in India, Bangladesh, Pakistan and Nepal this year, according to the United Nations.
With its people dying and floodwaters difficult to monitor, the Indian Space Research Organisation (ISRO) sent out a worldwide appeal Aug. 6 for help in tracking the swollen rivers of its Uttar Pradesh and Bihar provinces. Soon enough, satellites from space agencies across the globe pulled their focus tight on India, generating up-to-date maps for relief workers.
In few other areas of earth observation is time more of the essence than the emergency mapping demanded by natural disasters, “which is why it’s important to reduce the maximum delay between accessing and processing data,” said Luigi Fusco, senior adviser at the European Space Agency’s Center for Earth Observation (ESRIN).
This need to quickly analyze and share the terabytes of data continuously streaming down from satellites is partially why ESRIN has undertaken a project to adapt its systems to the grid, a technology that allows processing chores to be broken down to hundreds or even thousands of computers spread across the world.
Part of this project is in the hands of Iguassu Software Systems, which recently became the first domestic company to win a competitive tender from the European Space Agency (ESA). The company is collaborating with a consortium led by the Italian firm Advanced Computer Systems (ACS).
“The point of the grid is that it lets you share data and computing infrastructure between users,” said Martin Pačes, Iguassu’s lead developer for the project. “Researchers will come with their own software, put it on the grid and have instant access to ESA’s data.”
Previously, researchers would have to prepare proposals and then ESRIN would download data onto tapes for mailing.
“In the fastest case, it takes a couple of hours to get that information,” Pačes said. “For researchers, it can be days or months. But with the grid, the data is already there and you can process it as you wish.”
Iguassu’s contract comes under ACS’s larger commission to develop advanced methods of “multitemporal” analysis, which will allow ESRIN to quickly sweep through satellite data and find matching sets of geography that have changed over time — a flooded river, for example, or rice crops as they are grown over a season.
Researchers from across Europe have already devised several algorithms that can make these calculations, with some human training. (Teaching the computer the difference between forests in Central Europe and Southern Europe, for example.) Pačes’ task is to adapt these algorithms and the online system they’ll be run on — called KEO — to the grid.
Shared load
Grid computing came out of research conducted at CERN, the world’s largest particle physics center, located on the border between France and Switzerland.
CERN developed the basics of the grid’s load-sharing middleware, which serves as a bridge between operating system and software. This middleware, which ESRIN has modified, will allow CERN to distribute the data-intensive task of parsing particle collisions to computers across Europe, whether they are fleets of desktop PCs or solitary supercomputers.
It didn’t take long for other research agencies to see the grid’s potential, said Petr Bareš, managing director of Iguassu.
“[ESRIN] caught on to this because earth observation processing requires massive processing power, especially pictures from radar sensors,” he said. “Because of this, 80 percent to 90 percent of all satellite pictures that have been taken aren’t processed.”
ESRIN already has several applications running on its own grid, called G-POD. For example, G-POD generates daily mosaics of Antarctica from data sent by ESA’s Envisat satellite.
But this just touches the surface of the grid’s potential, and Pačes’ work creating ties between the grid and KEO could help speed this interaction.
“Applications like global mapping or sophisticated data mining that are run on KEO could really profit from using the grid,” Fusco said.
Emergency response
Further adoption of grid computing could speed response times to disasters, Fusco said.
Currently, countries send out appeals for emergency mapping through the Charter for Disaster Management, first signed between the Canadian Space Agency and ESA in 2000 and subsequently joined by other countries.
“Once the space agencies get a call to respond to the tsunami catastrophe or the dozens of floods that happened last month, they select what data they have,” Pačes said. “Many satellites are flying around, and sometimes they catch data, sometimes they don’t. At ESRIN, someone has to react quickly and collate this data.”
Some technologies developed by ESRIN have already sped up the agency’s response to disasters, Pačes said. KEO in particular has allowed much quicker interactions with satellite data.
“It seems like it’s helping a lot,” Pačes said. “Before, this all had to be done manually.”
Fusco would like to see the agencies go further, though.
“[The charter] would operate better if an infrastructure like G-POD could be shared across the space agencies participating,” he said.
Paul Voosen can be reached at pvoosen@praguepost.com
