 |
 |
|
DG Team Explores Methods of Detecting Crisis Points
But just like physical tools, algorithms and statistical models can be in need of updating. Data analysis techniques that once seemed sufficient can prove inadequate to the task of deriving answers from more complicated data sets with increasing numbers of parameters. "We need to develop a new kind of statistics and data analytics to deal with modern challenges," says Digital Government researcher Ganapati P. (GP) Patil, director of the Center for Statistical Ecology and Environmental Statistics at Pennsylvania State University. Patil and his colleagues are working on how to identify potential "hotspots." These can be crisis points in the military sense involving areas and networks of robots, sensors, or wireless devices, but also in the environmental and human health sense involving events of societal importance over geographic regions or across networks. Where might the next outbreak of West Nile Virus be? Where could an insect pest or invasive plant species overtake native agriculture and nature? Many of these concerns, which used to be of primary interest only to public health and environmental officials, now have also become issues for Homeland Security professionals. Where might prevailing winds take a bioterror agent? What is the most vulnerable spot downstream from contaminated and contaminating water sources upstream? The identification process for these hotspots requires the coordination of many spatial and temporal parameters. Consider, as one example, a potential epidemic. One would need to include geographic distribution of reported cases of an illness, to get an idea both of where it started and of whether its spread is following any particular pattern. One would also need to consider if there were unusual spikes in the numbers of cases in a brief period of time - and one would need to compare those figures to historical records to see if the number of cases is truly high in comparison to the same time period in previous years. In order to correctly incorporate and interpret all of these parameters, Patil and his collaborators start with the spatial scan statistic, a widely used metric in public health, and modify it to apply to environmental sciences and elsewhere. "The popular health-area scan statistic is a statistical method designed to detect a local excess of events and to test if such an excess can reasonably have occurred by chance," explains Patil. "However, its major limitation is that it is circle-based. The clusters can be of any shape, and cannot be captured only by circles. In more general settings, this is likely to give more false alarms and more of a false sense of security. What we need is capability to detect arbitrarily shaped hotspots. We plan to accomplish this using our innovation with upper level sets and their connected components."
What Patil and his collaborators are creating is, in his words, "the methodological toolbox" - the mathematical concepts that will, in a future phase of development, become the basis for prototype software that can be used by public officials and concerned scientists to anticipate and address crises ranging from natural diseases to homeland security events. "If you were to take slices of hot spots every day, you could see a trajectory that would give you an idea about the nature of the hotspots to target," says Patil. "The work that Patil is doing has opportunities to benefit everything from the Humanitarian Information Unit in the Department of State to ecological studies," says John Kelmelis, Senior Counselor for Earth Science at the U.S. Department of State, "The [HIU] responds to complex humanitarian disasters worldwide - identifying them before they happen is very important. Patil's work will help us do that. Additionally, it will be helpful in identifying fundamental changes in earth processes, like land cover changes. Once you've been able to identify them, then you can work on determining what caused them." |
||||||||
 |
This site is maintained by the Digital Government Research Center at the University of Southern California's Information Sciences Institute. |  |
" CONTACT " POLICIES | ||
| |
|||||
