Public Release: 

ORNL program could help city dwellers survive disaster

DOE/Oak Ridge National Laboratory

OAK RIDGE, Tenn., Aug. 21, 2002 - Most population databases have one potentially fatal flaw, according to Oak Ridge National Laboratory researchers who have developed a system that could be a lifesaver if a city had to be evacuated.

Unlike conventional models, ORNL's LandScan USA model takes into account the difference between daytime and nighttime populations in cities across the country. The distinction is an important one but is ignored by models that rely solely on U.S. Census Bureau data that represent nighttime or residential population.

In Houston, for example, about 160,000 people move into the downtown business district every weekday and return to their suburban homes each evening. That number isn't included in census information. And in Chicago, the Sears Tower has zero population, according to census data.

"If you're a terrorist, you can round to tens, hundreds or thousands, but if you're trying to save lives, each and every person matters," said Budhendra Bhaduri of ORNL's Computational Sciences and Engineering Division.

Another flaw of traditional population distribution models is that, even at the highest resolution -- the block level - they assume a uniform distribution of people. This is one of the weakest assumptions, Bhaduri said, because people rarely live uniformly even within a city block.

"It's apparent that to enhance our capability in emergency planning and management, we not only need to estimate population at an even finer level, but we also need to have a fairly reasonable estimate of population distribution during the day," Bhaduri said.

In ORNL's global population projects, Bhaduri's group is using an innovative approach with geographic information system and remote sensing to develop a high-resolution population distribution model (LandScan) for the world. At a 1-square-kilometer resolution, LandScan is the finest global population data ever produced and is 2,400 times more spatially refined than the previous standard.

"Hundreds of professionals at government agencies, educational institutes and research organizations around the world are taking advantage of the improved population distribution estimates provided by LandScan to facilitate decision-making processes that rely on population data," said Eddie Bright, who leads the global LandScan development.

For the United States, ORNL researchers are in the process of refining that even further to allow for resolution down to a 90-meter cell, or smaller than most city blocks.

At the 90-meter resolution, the data includes daytime and nighttime - or residential - distributions. Researchers have demonstrated the potential benefit of LandScan USA in 29 counties covering coastal Texas and Louisiana - including the Houston metropolitan area.

In addition to its application for emergency planning in case of an attack or natural disaster, LandScan has potential uses for socio-environmental studies, exposure and health risk assessment and urban sprawl estimates. The model also can be modified for other applications and is being considered as the primary tool to estimate spatial distribution of pesticide usage in urban watersheds.

To develop the model, researchers divide census blocks into finer grid cells and evaluate each for the likelihood of being populated. They base this on relevant spatial characteristics, including land cover, slope, proximity to roads and nighttime lights. Researchers then allocate the total population for that block to each cell taking into account the calculated likelihood, or population coefficient, of being populated.


Funding for LandScan USA has been provided by DOE, the Department of Defense and the Environmental Protection Agency. ORNL is a DOE multiprogram research facility managed by UT-Battelle.

You may read other press releases from Oak Ridge National Laboratory or learn more about the lab at

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.