Traverse City Record-Eagle

Latest News - Mobile

December 8, 2013

Scientists turn to math to analyze, predict weather

LANSING — Think all ice is the same?

That’s not the case on Michigan’s Great Lakes. And now scientists have found out how to detect the differences — with math.

The development, reported recently in the International Association for Great Lakes Research, is important because it could help guide freighters through Michigan’s icy lakes, assist the Coast Guard in breaking up large ice formations and help weather scientists predict evaporation that could lead to lake effect snow.

Researcher George Leshkevich of the National Oceanic and Atmospheric Administration in Ann Arbor led the study with help from Son Nghiem of the National Aeronautics and Space Administration. The two developed a radar system and an algorithm to detect types of ice formations on the Great Lakes.

An algorithm is set of steps that are followed to solve a mathematical problem.

The researchers use radar systems from satellites or mounted locations to bounce a signal off the ice.

The radar sends back what Nghiem and Leshkevich call a “signature,” which can be interpreted using an equation to determine which type of ice lays on the water surface. This signature and resulting information from the equation allow scientists to measure things such as density and depth of ice.

Just as the handwriting of two people will never look exactly the same, different types of ice will return a unique signature that scientists can decipher.

Leshkevich created an algorithm that can differentiate one ice type from another by analyzing data from as far back as 1997. It allows researchers to differentiate among ice types by plugging in the data returned to the radar mechanism, and reading it to see which ice lies below.

It produces data that’s specific to each kind of ice formation, said Nghiem.

“You kind of create a dictionary which can translate the radar signature into a physical environment, like a different kind of ice type,” Nghiem said.

Text Only