The Institute of Technology is in a strong position to help tackle many of the most vexing scientific questions about climate change
Vipin Kumar: Modeling the impact
by Richard Broderick
“We are computer scientists,” declares Vipin
Kumar, head of the Department of Computer Science
and Engineering. “We develop algorithms and software
that, among other things, can be used for data
mining of complex information.”
 |
| Vipin Kumar, head of the department of computer science and
engineering, and graduate student Shyam Boriah, are developing
algorithms and software that are being used by NASA to mine
satellite images that create a history of changes in Earth’s landcover. |
Right now—and most critically when it comes to
climate change—the data mining algorithms and
software developed by Kumar and Shyam Boriah, a
graduate student in computer science and engineering,
are being used by NASA to mine satellite images
to create a history of changes in Earth’s land-cover.
In particular, detecting changes in the forest ecosystem
and its recovery period is critical for sustainable
management of forest resources, monitoring the impacts
of climate change on forests, documenting a
nation’s compliance with United Nations protocols,
and carbon trading.
Kumar and his team have also developed algorithms
to identify sudden changes in Earth’s forest
cover—a sure signal of rapid deforestation, which
can contribute a double-whammy of CO2 to the atmosphere.
For example, fire releases carbon dioxide
directly while fewer trees mean fewer leaves to absorb
CO2.
“There’s a lot of satellite data now going back to the
1980s,” Kumar explains. “It’s a painstaking process to
sort through that data in any meaningful way to determine
what changes have taken place over time.”
In a nutshell, that’s what Kumar’s spatio-temporal
data mining techniques are designed to do.
They also employ data collected by satellites to map
out the interactions between ocean temperatures,
weather patterns, and land cover in the larger pattern
of fluctuation of atmospheric CO2 levels.
“We have developed algorithms that take data
from different parts of the globe and use them to
identify changes to the Earth’s surface,” he said.
“Questions about how land use is changing are important
for many reasons, but especially for policy
planning.”
One of the first applications of Kumar’s data mining
system for detecting land cover change was done
for the state of California. “When we applied it to
the Bay Area vegetation data, we were able to detect
changes like farmland being converted to housing
subdivisions and golf courses,” he said. “When we
applied our software to the state as a whole, it identified
numerous conversions of desert into farmlands
and a large number of forest fires.”
Once the systems Kumar is developing are completely
fleshed out, he sees a number of potential
uses beyond environmental policy.
 “The biggest use, of course, will be for carbon
modeling—the folks who want to know where vegetation
is changing so they can use that information
to help build models of how much carbon is going
into the atmosphere and how quickly,” he said. In
this regard, it’s vital to know the time frame over
which change has occurred.
“It makes a big difference if trees are burned down
or if they are used to make furniture,” he said. “It’s
quick versus slow release of carbon. To do precise
carbon cycle modeling, we need to know the answer
to these kinds of questions.”
But Kumar sees one additional role for his team’s
algorithms and software—public education.
“Our goal is to make this system available to the
public so that even schoolchildren will be able to use
it and see what is happening to our environment.”
|
