“We were able to use the information in a good way to save taxpayers money rather than haphazardly throwing things against the wall and seeing if it sticks,” said Virginia State Police Capt. Steven W. Lambert of the Virginia Fusion Center, which coordinates statewide investigations, including the case of the five shootings at military buildings.
When Yonathan Melaku was arrested at Arlington National Cemetery in summer 2011, months after he started his string of shootings, police had been expecting him there.
They had analyzed hundreds of factors, such as the gunman’s sightlines, access points and escape routes. His escape route was particularly important; he stayed close to highways.
With the data, they produced a list of hundreds of likely target areas, along with the probability that the shooter would visit each one.
It helped them figure out where to focus resources and what to protect.
It didn’t lead to Melaku’s arrest; he was stopped in the cemetery behaving erratically and lugging a backpack loaded with spray paint, spent shell casings and homemade bomb-making materials.
The case, led by the FBI, ended with Melaku pleading guilty to the shootings, and in January he was sentenced to 25 years in prison.
But the accuracy of the prediction — placing Arlington in the top tier of his possible targets — seemed to prove the technology’s potential in helping to focus large, complex investigations, Lambert said.
“This helps us make sense of our data,” Lambert said. “In the Fusion Center, I often say we’re drowning in information and yet starved for knowledge.”
Other data police analyzed included terrain of the crime scenes, public transit access and the economics of surrounding neighborhoods.
The technology initially was developed using satellite images and other data to help U.S. troops anticipate where explosives were buried in Iraq and Afghanistan. It is one in a string of wartime science applications — high-speed license plate readers, mobile fingerprint devices, facial recognition software — being used by police in local crime-fighting.
Around for decades
Predictive technology has been around for decades, going back to when the U.S. Air Force needed to anticipate when its machines and equipment would need replacement.
In an updated form, it is similar to what retailers such as Amazon and Wal-Mart use to anticipate buyer behavior, said Colleen McCue of DigitalGlobe, the private company that developed the technology and worked with law enforcement on the Melaku case.
The theory is that people, including criminals, tend to go where they are comfortable. The key is identifying those patterns because they will probably be repeated.
“People are creatures of habit,” said McCue, who specializes in behavioral analysis of violent crime. “It’s not a coincidence that in the grocery store your favorite cereal is at eye level.”
In Iraq and Afghanistan, DigitalGlobe used satellite images to track insurgent settlements and discovered that explosive sites were often near walled compounds that provided cover or trails between Afghan vineyards that could be used for discreet travel.
Less likely were sites near fallow vineyards, where movements were more exposed.
Similar policing tactics are being used by the FBI and police departments across the country to get ahead of crime, including in the District, New York and New Jersey. The Virginia State Police recently used it to investigate copper thefts at Dominion Virginia Power facilities, finding that one of the key predictors of which site would be hit next was how close the sites were to scrap-metal dealers where the copper could be sold.
Police say they are employing the technology as they hunt for an arsonist on Virginia’s Eastern Shore blamed for 48 fires since November.
New Jersey’s Regional Operations Intelligence Center used it in 2010 when police were tracking a serial strangler in Philadelphia who killed three women. DNA ended up linking Antonio Rodriguez to the crimes, and he was convicted of murder last summer.
“It’s the future of law enforcement,” said Lambert, the Virginia investigator.
In the Melaku case in Virginia, the initial area thought to be in his target zone was more than 750 square miles.
After the analysis — which included three-dimensional views of crime scenes and surrounding areas — it was condensed to 75 square miles, with a focus on much smaller hot spots.
“Think about the Northern Virginia landscape,” McCue said. “You can’t throw a rock without hitting a national security target. It was an enormous area that spanned a couple of states.”
Melaku started the spree in October 2010 by twice shooting at the Marine Corps museum in Triangle, 36 miles south of Washington.
Later that month, he upped his visibility by shooting windows at the Pentagon.
Then he fired on a Marine Corps recruiting substation in Chantilly.
Of particular concern to investigators was the fast-approaching Marine Corps Marathon.
McCue helped develop a heat map using dozens of data points from the first four shootings. The result showed likely future target areas, enabling police to adjust security and deployment.
In November, Melaku initially threw investigators for a loop by attacking a facility that was not affiliated with the Marines — a Coast Guard recruiting center in Woodbridge.
But when investigators went back to the heat map, they found that the Coast Guard center had been identified as a “high likelihood” spot on the map, confirming the analytics.
Then, after a long pause in the shootings, Melaku turned up at Arlington National Cemetery. He had caused more than $110,000 in damage to military buildings.
“At end of the day, nobody got hurt,” McCue said.
McCue’s work with the Virginia Fusion Center was funded by a grant from the Department of Homeland Security, which is exploring new applications for the wartime technology.
Another group called PredPol was started by California researchers who helped a police department in Santa Cruz predict crime to guide the department’s dwindling resources.
“In the last decade, we had a 30 percent increase in calls for service and a 20 percent decline in staff,” said Zach Friend, the department’s crime analyst at the time, who is now a member of the county’s board of supervisors. “We knew we weren’t going to get more cops on the street.”
At the beginning of beat officers’ shifts, they now are given a map detailing which crimes might occur — burglaries, car break-ins, thefts — within an identified zone.
In the first six months of the program, burglaries declined 19 percent.