When Absalom Jordan hears the crack of gunfire outside his home in Southeast Washington, he reacts in an instant. “You get away from the windows and get down,” the 72-year-old said. “I have learned to live with it.”
Police are listening as well. Rooftop sensors monitor his neighborhood around the clock for the distinctive bang of a gun. The inconspicuous devices have logged hundreds of incidents over the past eight years near his apartment as part of a gunfire surveillance network called ShotSpotter.
About 39,000 separate incidents of gunfire have been documented by ShotSpotter’s unseen web of at least 300 acoustic sensors across 20 square miles of the city, according to an analysis by The Washington Post. The data, obtained through a public-records request, offer an unprecedented view of gun crime in a city where shooting a firearm is illegal in virtually all circumstances.
The gunfire logged by ShotSpotter overshadows the number of officially reported felony gun crimes by more than 2 to 1. More than one-half of the incidents detected by the network have involved multiple rounds of gunfire. In 2009 alone, ShotSpotter captured more than 9,000 incidents of gunfire. That number has fallen by 40 percent in recent years as gun homicides have declined sharply.
The system has been helpful to law enforcement, but no one claims that it captures every shot. The network covers only a third of the city, focusing on the police districts with the most violent crime. It occasionally misses gunfire because of circumstances that can cloak acoustic signatures, such as the canyonlike structures of an urban landscape. Some sounds, such as fireworks, can be mistaken for gunfire, although technology and human review help weed out false positives.
GUNS IN AMERICA: This article is part of an occasional series examining the role of guns and the influence of the gun industry in the United States.
When ShotSpotter’s remote monitors — microphones and circuitry in a weatherproof shell — detect a loud noise, a central computer program analyzes the acoustic signature, providing a more accurate location than people usually can. It classifies the source, pinpoints the suspected location to within a few yards and notifies police. City personnel verify the alert and dispatch officers.
“ShotSpotter gives you a specific location,” said Kristopher Baumann, president of the D.C. police union. “You can go there and get out of the car. You can find a victim or shell casings.”
ShotSpotter’s coverage is most extensive in the eastern half of the city. By quadrant, the network has captured 18,700 incidents in Southeast, 10,600 incidents in Northeast, 6,400 in Northwest and 1,600 in Southwest, which is primarily waterfront and contains large stretches of undeveloped industrial areas. The network has logged an additional 1,600 shootings along the edges of the city.
Weather appeared to influence the pattern. The month of February had the fewest incidents, and July had the most. Work schedules also played a role. By day of the week, Saturday, Sunday and Friday had the most and Wednesday the fewest. As did sleep: The quietest hour was from 7 a.m. to 8 a.m.
A disproportionate number of incidents, about a third, have been logged on and around New Year’s Day and July 4. Much of this, officials said, is likely the result of celebratory gunplay.
The District is the biggest client of SST of Newark, Calif., which produces ShotSpotter. It’s also the only city in the region to use the system, although Prince George’s County police are testing it. Other companies provide listening technology for gunfire, but SST says it is the only one that can pinpoint shots over wide areas.
District officials praise the system.
“It is a valuable tool that provides almost instantaneous alerts that allow officers to be dispatched quicker for the sound of gunshots,” Police Chief Cathy L. Lanier said in a written statement. “It has also been instrumental in determining crime trends and establishing information in investigations.”
The system can help police identify turf battles between gangs or other gun-related crime affecting particular neighborhoods, said Cmdr. James Crane, ShotSpotter program manager.
The gunshot detection technology, advocates said, also helps law enforcement address two problems: People misidentify sounds, such as cars backfiring, as gunshots, and true gunfire often goes unreported. ShotSpotter officials said that studies among clients elsewhere have found that about four out of five gunfire incidents are never reported to police.
The company guarantees that the system will capture at least 80 percent of all audible, outdoor gunfire in coverage zones, but company officials said they typically achieve a rate of 90 to 95 percent.
ShotSpotter’s inventory of shooting incidents provides another measure of gun-related crime in the District. In May, The Post reported that more than 28,000 firearms, mostly handguns, have been confiscated by D.C. police since 2000, though in smaller numbers in recent years.
The gunfire documented by ShotSpotter — an average of 17 incidents a day in the District since 2009 — offers a fuller view of the “pyramid” of gun violence, said Daniel Webster, who directs the Johns Hopkins Center for Gun Policy and Research in Baltimore.
“At the top, you have the incidents when someone dies. Then you have those when somebody gets wounded,” Webster explained.
The biggest category of gun violence, he said, is also the most underreported and poorly documented: when a shooter misses the target or shoots to intimidate.
“It gives a much better picture of how prevalent gun violence is,” Webster said.
ShotSpotter grew out of one man’s concern about gun violence on the West Coast in the early 1990s.
An engineer and expert in acoustic sciences, Robert Showen, then in his early 50s, was working at a research institute in Menlo Park, Calif. Showen said he was troubled by the deadly gang wars in nearby East Palo Alto.
“I thought, with my knowledge I can do something,” Showen said.
He said he asked local police whether a system to detect gunfire could help fight crime. They directed him to scientists from the U.S. Geological Survey who were working to apply earthquake location systems to gunfire. Their research showed promise, Showen said, but they had not figured out how to apply it to an urban environment in real time.
Showen experimented: He set microphones atop a piano in his living room and connected them to a laptop outfitted with software that he and a business partner wrote. Using a child’s clicker toy, he made a noise to see whether the software could detect where the sound originated. He conducted the same test outdoors, popping balloons at various spots in his yard. His system worked.
In the mid-1990s, he took his idea to police in nearby Redwood City, Calif., who became the first to deploy ShotSpotter. Contracts soon followed in Los Angeles County and Glendale, Ariz. Showen is now chief scientist for SST and co-holds a patent on the technology. Today, 65 police agencies use ShotSpotter in the United States, as do police in Rio de Janeiro. It’s also being tested in a South African safari park to see whether it can help identify illegal rhinoceros poaching.
The District began using ShotSpotter in late 2005 after the FBI gave the city the opportunity to test the technology. A federal grant paid $2 million to place sensors across the 7th Police District in Southeast Washington.
“The 7th District was selected for ShotSpotter because it led the city in homicides,” said Joel Maupin, who was police commander of the district and has since retired.
The technology enabled officers to respond to shootings more quickly. It became a safety issue. In short order, the department had to revise its dispatch policy to ensure that officers knew whether the report of gunfire came from ShotSpotter or from a person who thought he heard a gunshot, said former D.C. police chief Charles Ramsey.
“Because the odds [with ShotSpotter] of finding a person armed was a lot higher than with a normal 911 call,” Ramsey said. “It could distinguish cars backfiring from gunshots.”
In 2007, the District assumed ownership of the detection system, expanding its coverage in the years that followed. ShotSpotter now reaches into six of the seven police districts and covers about one-third of the city. Its greatest coverage is in Southeast and Northeast, records show.
Over the past six years, the city has spent about $3.5 million to maintain and expand the system, records show.
ShotSpotter is also linked to a system of closed-circuit cameras, which police hope will capture the aftermath of shootings in real time. To guard against vandalism, officials do not publicize the sensors’ appearance or reveal their locations.
Wrapped in a weatherproof container roughly the size of a watermelon, each ShotSpotter sensor combines microphones, hardware, software and a clock linked to the Global Positioning System, which uses satellites and radio navigation to pinpoint precise times and locations.
In the cacophonous urban environment, sensors are calibrated to ignore all sounds except for those that most closely match the “impulsive” sound of an explosion, said James Beldock, a senior vice president for ShotSpotter.
“It’s a very, very sharp wave,” Beldock said. “No other sound works that way.”
The blast of a gun is different from other explosive sounds because it is directional, meaning that the noise changes its frequency as the bullet moves through space. A person may hear a gunshot a half-mile away if the gun is fired toward him. But a person 200 yards away may hear nothing if the gun is fired away from him.
Once sensors register a potential gunshot, they transmit the data to the ShotSpotter computer network for analysis. The computer server compares the time that each sensor logged the sound to calculate the likely location of its source, a process of triangulation and multilateration.
“That sound will reach a sensor 100 yards away at a different time than it reaches a sensor 200 yards away,” Beldock explained.
The more sensors that capture the noise, the more accurate the location. A sound detected by 10 sensors can be located to within two feet, he said.
The computer system also classifies the likely source of the sound based on its sharpness, frequency and consistency across sensors. This is critical, because other impulsive sounds — including fireworks, backfires and helicopters — can also trigger the remote sensors.
Pile drivers, for example, initiate an alert because the machinery is elevated and the noise radiates over a long distance, Beldock said. But it can often be filtered out.
“The pile driver does not sound like gunfire in one critical respect: The frequency component, the pitch of the sound is not right,” Beldock said.
The software will try to determine whether the source of the gunfire was in motion and might have been a drive-by.
Can you tell the difference?
Each of these three sounds is either a gunshot or a pyrotechnic. Find out more about how ShotSpotter works.
1 of 3
MOUSEOVER TO LISTEN
2 of 3
MOUSEOVER TO LISTEN
3 of 3
MOUSEOVER TO LISTEN
After the network classifies the sound, a person will review it. “The job of the review team is to use their knowledge of other things that the computer is not yet good enough to do,” he said.
A reviewer will listen to the sound and visually inspect its wave on a monitor. A gun blast looks like a Christmas tree tipped over on its right side: “The bushier the tree, the more likely it is to be a gunshot,” he said.
Fireworks are among the most difficult sounds to discern, but they can often be identified because, unlike gunfire, the intensity of the explosion is the same in all directions. Some of these invariably slip through. In April, ShotSpotter logged 39 gunfire incidents in the middle of the Washington Channel during the Cherry Blossom Festival fireworks display.
A reviewer will also try to determine whether the gunfire came from an automatic weapon and whether more than one gun was fired. “The musical analogy is that if the rhythm is not completely even, then it’s likely to be multiple shooters,” Beldock said.
In the District, city personnel review alerts. For many other cities using a different version of the ShotSpotter network, that work is done by company personnel in a command center in Newark, Calif. They pass additional information gleaned from maps or satellite photos to local dispatchers.
The entire process of detection through review for ShotSpotter personnel takes less than 40 seconds, on average. Once a gunshot is verified, police are dispatched. ShotSpotter officials call this “get the cop to the dot.”
D.C. police declined requests to demonstrate the system. Since its installation, police have not publicized the amount of gunfire captured by the network in their annual reports. And they said they do not track arrests made as a result of ShotSpotter’s alerts.
Nationwide, criminologists say they are not aware of recent academic research on the effectiveness of such systems. A 2011 study commissioned by the company found that seven agencies using its technology reported better shooting response times and more efficient investigations.
For investigators in the District, ShotSpotter played a key role in a high-profile 2007 shooting in which an off-duty D.C. police officer killed a 14-year-old boy on a stolen minibike. ShotSpotter’s audio captured gunfire from two sources, evidence that police said indicated that someone other than the officer fired first. That conclusion provided a strong defense for the officer. A department review said the police use of force was justified and that the officer did not violate department policy. The family sued the District, which ultimately paid a settlement to end the case.
ShotSpotter information is “not frequently used at trials” but has helped prosecutors establish the number or sequence of shots, the time of gunfire and whether more than one gun was fired, said William Miller, spokesman for the U.S. attorney’s office.
Prosecutors said the technology helped win a voluntary manslaughter conviction in the slaying of 20-year-old Deuante Ray in the early morning of Oct. 30, 2009. Prosecutors called it a “difficult circumstantial evidence case.”
A gunman shot Ray in the head and chest, killing him in an alley behind the 1100 block of 48th Street NE. A witness last saw Ray entering the alley that night with Terrell Patton. Patton denied shooting Ray but admitted that he lent Ray his cellphone.
Records showed that two calls were made from the phone that night. The second call by Ray, to his girlfriend, ended about 12:35 a.m. ShotSpotter detected a gunshot in the alley about 15 seconds after the call ended. That, prosecutors said, was one of the shots that killed Ray. Patton, they said, lured Ray into the alley and shot him.
“Technology provided key evidence,” prosecutors noted.
The thousands of incidents logged by ShotSpotter stress one fact: Many D.C. residents can’t escape the crack of gunshots, though they have become less frequent, falling to 5,385 in 2012. The Post plotted eight years of incidents on a map of the city, using the latitude and longitude of each.
A cloud of virtual gunfire emerges, flecked by hot spots.
In Northeast, the neighborhood surrounding Clay Terrace has been subjected to 302 incidents in the five years since 2008. In 2009, the number peaked at 129. Last year, it dwindled to 17. The area is home to a public housing complex flanked by the Arts and Technology Academy and Marvin Gaye Park.
Greg Stewart, who lives nearby and serves on the local Advisory Neighborhood Commission, noted the change.
“Right now, you don’t really hear complaints,” said Stewart, who said he was aware of ShotSpotter from news reports and had noticed an increased police presence in the area.
In Northwest, where ShotSpotter’s coverage is more limited, a portion of the Columbia Heights neighborhood east of Georgia Avenue between Princeton Place and Lamont Street experienced 299 incidents over the past five years. The number peaked at 99 in 2009, dropping to 46 last year.
Southeast has the greatest number of hot spots. It also has the greatest ShotSpotter coverage.
ShotSpotter has captured 329 incidents in a part of the Washington Highlands neighborhood near the 4300 block of Fourth Street SE, just north of the Maryland border.
“There were times when I was awakened by shots. Shootings are a constant thing in my area,” said Jordan, the 72-year-old who lives there and is a member of the local Advisory Neighborhood Commission. In August, a man was shot to death in front of a nearby apartment.
A few miles away in Southeast, ShotSpotter has recorded a surge in shootings this year along Bruce Place, a short residential street on the edge of Fort Stanton Park. In the first half of the year, sensors had logged 63 incidents, 22 of them involving multiple rounds of gunfire. In all of last year there were only 45.
“The amount of gunfire in this area highlights the challenges we are facing and addressing,” D.C. police Cmdr. Robin Hoey said. Police did not provide an explanation for the increase.
Over the eight years of its operation in the District, ShotSpotter has logged more than 50 incidents each within a quarter mile of a dozen schools during weekdays.
In Northeast, for example, ShotSpotter recorded 94 shootings within a quarter mile of the Friendship Blow Pierce Junior Academy on 19th Street since 2008. The shootings occurred between 6 a.m. and 6 p.m. on school days.
“We hear about the incidents. But when we are inside the school, we don’t have a sense of violence,” said Patricia Brantley, the chief operating officer of Friendship Public Charter School, which runs the Blow Pierce Junior Academy.
Some gunfire still escapes ShotSpotter’s net.
The Post identified about a dozen gun homicides that took place outdoors in the coverage area that the system failed to detect. ShotSpotter officials and police confirmed the findings.
On Clay Terrace NE, ShotSpotter missed a high-profile shootout on Oct. 13, 2009, that left two dead and three wounded. Gunmen opened fire shortly before 4 p.m. on a group of people standing in a courtyard between two buildings. Some in the crowd drew guns and returned fire, records show.
“I heard about 30 gunshots, and it sounded like three different guns,” Darryl Profit, who was inside his house a half-block away, told The Post at the time. Another witness likened the fusillade to a “war.” Police at a school nearby also heard the gunshots.
When the system appears to miss a shooting, “there is a very formal and rigorous review,” said Crane, the ShotSpotter manager for D.C. police.
Officials have worked with ShotSpotter to fill any gaps in coverage, installing additional sensors as needed.
Certain circumstances can thwart detection, officials said.
If a silencer is used or shots are fired into a car and the vehicle absorbs the acoustic energy of the blast, the sound may elude the sensors, said Ralph Clark, ShotSpotter’s chief executive. The same is true, he said, for an “execution style” shooting in which the gun is inches away from the victim. He also said that gunfire in a canyonlike area could be clouded by ambient noise.
“We don’t offer a 100 percent ironclad guarantee to capture 100 percent of all the shootings,” Clark said. “What we provide is a lot of gunshot intelligence that otherwise would not be attainable by any agencies.”
Ted Mellnik and Clarence Williams contributed to this report. Petho is an investigative journalist from Hungary who is reporting for The Post under the Hubert H. Humphrey Fellowship Program.