Editor’s note: This is one of a series of five posts profiling finalists for NVIDIA’s 2015 Global Impact Award, which provides $150,000 to researchers using NVIDIA technology for groundbreaking work that addresses social, humanitarian and environmental problems.
Refugee camps in West Africa. Mobile homes in U.S. tornado corridors. Both densely populated. Both with different needs.
How should governments, health agencies and first responders allocate resources to such areas with fast-changing circumstances? One way would be to learn more by knocking on doors, but that’s impossibly slow and not feasible in many parts of the world.
So some geographers, computer scientists and engineers are instead deploying the world’s fastest supercomputers to map and analyze population size and shifts in unprecedented detail.
To do this, researchers at Oak Ridge National Laboratory are using NVIDIA GPUs to produce LandScan high-definition global population data.
“Every time you create services, you have to understand where the people are. And when you are trying to mitigate risks, every life counts,” said Budhendra Bhaduri, who leads the Geographic Information Science and Technology group and the Urban Dynamics Institute at Oak Ridge.
The work has placed Oak Ridge among five finalists for NVIDIA’s 2015 Global Impact Award. We award our annual grant of $150,000 to researchers who use NVIDIA technology for groundbreaking work addressing social, humanitarian and environmental problems.
Transforming Urban Planning
Oak Ridge’s analysis is transforming urban infrastructure planning.
It helps guide where to build schools and hospitals. It helps pinpoint areas in critical need after natural disasters.
Knowledge gleaned from polio eradication efforts was even applied to the emergency response to the West African Ebola outbreak.
Top-down modeling from national censuses and satellite images are the traditional tools. But “lots of countries below the equator don’t have a regular census,” said Bhaduri.
Oak Ridge researchers have taken a novel approach, infusing technology into urban planning to measure populations.
“We thought, can you go bottom-up and rewrite population assessment without a national census?” Bhaduri said. “We asked: How can we understand where people are? Then GPUs changed the whole game.”
With the introduction of very high-resolution satellite images, the mapping of smaller settlements became possible on a global scale. This is particularly helpful for remote regions in many less developed countries.
With the research and data analysis, “we could come up with a completely independent estimate of how many people are really on this planet,” Bhaduri. “In some terrains we’re mapping people for the first time in human history.”
Mapping settlements involves advanced algorithms capable of extracting, representing, modeling and interpreting satellite image features.
A decade ago, automated feature extraction algorithms on CPU-based architectures helped speed the identification of settlements. But identifying quick shifts in population—such as migration or changes after a natural disaster—required more computing power.
The parallel-processing capability of NVIDIA Tesla GPUs allowed researchers to develop and use the expensive feature descriptor algorithms to process imagery at dramatic speed-ups of up to 200x.
“GPUs reduced the time to hours rather than days—on regular hardware it would take forever,” said Dilip Patlolla, a research scientist at Oak Ridge.
“With GPUs, you can analyze datasets in ways we never could before,” Patlolla said. “You can periodically map a refugee camp to estimate the influx of refugees—that helps planning for humanitarian assistance.”
“Similar to refugee camps, temporary settlements such as mobile homes in the U.S. are also being detected since they are highly vulnerable to natural hazards such as tornadoes,” he said. “It comes down to saving lives.”