Melissa Hanham, Jeffrey Lewis (left) and Shea Cotton (right) listen as Dave Schmerler (center) — a researcher nicknamed “Geolocation Jesus” by Lewis, who runs the center’s east Asia program — uses satellite images and 3-D models to locate the sites of North Korea’s missile tests and nuclear showrooms. (Anna Fifield/The Washington Post)

There were reports going around last month that North Korea had tested another solid-fuel missile engine, a type of engine that can be deployed much faster than the older liquid-fueled ones. 

Kim Jong Un’s media outlets hadn’t bragged about it — as they had done in previous tests — so the experts at the Middlebury Institute of International Studies’ nonproliferation center got to work.

They figured that the North Korean rocket scientists would have used the same immovable concrete block they used for an engine test last year.

Dave Schmerler — a researcher nicknamed “Geolocation Jesus” by Jeffrey Lewis, who runs the center’s East Asia program — had quickly located the site of the earlier test. 

He’d made 3-D models of the buildings in the North Korean photos and noted the surroundings. Then he’d taken official reports about Kim’s recent public activities — in that case, the leader had just been to a machinery plant near Hamhung on the east coast — and wham, he pinpointed the exact building on Google Earth.

Technology is making it possible for open-source analysts to do the kind of work previously the preserve of intelligence agencies and, in the process, learn all sorts of things about one of the most impenetrable countries in the world — one that wants to send nuclear-tipped missiles to the United States.

In addition to Lewis’s team in Monterey, satellite imagery analysts Curtis Melvin and Joseph Bermudez are publishing their findings on the website 38 North, run by the U.S.-Korea Institute at SAIS.

“Back in the day, if the government told you something, you had to believe it. That’s how we got the Iraq War,” said the voluble Lewis. “Our animating principle is that having a robust public debate about nuclear and missile technology in other countries is going to lead to better policies.”

Now, with unconfirmed reports of another test, Lewis and ­Schmerler and their colleagues set out to confirm them. 


Two images show a test stand location near Hamhung, North Korea. The image on the left shows the test stand on Oct. 7. The image on the right, from Oct. 17, shows vegetation near the test stand has changed between the time of the reported test. Researchers at the Middlebury Institute of International Studies use this near-infrared imagery from Planet Labs to confirm reports of missile tests. (Planet Labs)

Using near-infrared imagery from Planet Labs, a San Francisco-based company that takes daily images of the Earth, they scoured the photos of the previous engine test site, taken over several days in the middle of October. Google Earth might have better-quality images, but they can be months old.

Enhancing the Planet images with near-infrared light, which can differentiate between living and dead vegetation, the experts found signs of significant heat and force going in one direction away from the test site. 

North Korea did, indeed, appear to have tested another rocket engine. 

Using these methods of “open-source intelligence,” Lewis’s team has been able to spot developments and sound the alarm about the rapid progress in North Korea’s missile program, while government officials and other analysts were saying Pyongyang was still years away from making a rocket that could reach the United States.

“I find it really reassuring that there are people who can do this in open source,” said Melissa Hanham, a senior researcher at the center.

In one case, in October last year, they used before-and-after satellite imagery enhanced with near-infrared light and determined that there was an even chance that North Korea had just tested part of an intercontinental ballistic missile. 

Eight months later, North Korea launched its first ICBM, one that could reach Alaska. A few weeks later, it fired one that had Denver and Chicago in range.

In September 2016, North Korea's state-run television confirmed the country conducted its fifth nuclear test. (Reuters)

Satellite photos provide the analysts with lots of useful material. “I stare at North Korea all day,” Schmerler said. “I look at what’s happening and where it’s happening and everything beyond.”

North Korea also provides the researchers with plenty of valuable data through photos and videos the regime publishes following successful missile and nuclear tests. 

“When we get pictures, we’re able to start working to determine what actually happened. And it especially helps when they release video,” said Shea Cotton, another researcher on the team. 

With video, software can measure how quickly the missile launches off the pad. With that information, analysts can calculate its thrust. 

In photos, joints and lines on the missiles can help them make precise measurements of their size. The Japanese crane used to lift an intermediate-range missile in May told them its maximum weight.

The color of the flames from the missile launches can reveal whether the engine is powered by liquid or solid fuel. Solid-fuel rockets are faster to launch and therefore harder to detect in advance.

Gleaning information about North Korea’s nuclear weapons is, however, much more difficult because warhead design is classified by nature.  

It’s impossible to tell whether the peanut-shaped two-stage device that North Korea showed off in September, just hours before it detonated what it said was a two-stage hydrogen bomb, was a mock-up of the actual bomb that was tested.

“We can’t prove whether it’s real or not. We don’t have X-ray vision,” Hanham said. “But they’re trying to tell us that this is something they’re working on.”

Still, Nate Taylor, a graduate student, spent hours making a 3-D computer model of the “peanut,” counting bolts and divots, adding the 12 detonator wires. But, annoyingly, the device was photographed at an angle, not straight on, making it more difficult to get an accurate measurement. 

Taylor worked with what he had: Kim Jong Un’s face. He figured out the size of Kim’s face by comparing it with the basketball player who’s visited him — “because Dennis Rodman is our known quantity” — and used those numbers to calculate the length and diameter of the bomb.

But this emerging field comes with its own problems. If the experts figure out North Korea’s mistakes, is it ethical to disclose them? “It’s cool to make the models, but we don’t want to help solve their problems for them,” Hanham said.

Although the bombs are difficult, Lewis’s team has been able to determine some information about where North Korea is showing them off — and there is some evidence that Pyongyang may be trying to thwart the team.

When Kim inspected the first warhead North Korea publicized — a round device nicknamed the “disco ball” — it didn’t take ­Schmerler long to find the building.

He counted the high beams on the roof and measured the skylight windows and went “cruising” over North Korea and found the exact building. Matching beams, matching skylights.

The next time North Korea published a photo of a warhead — the “peanut” detonated in September — the photos were taken in a completely white room where there was nothing to measure. No beams, no windows, just sloped floors that suggested it was underground. 

“Hats off,” said Schmerler (although he thinks he still figured out the location).


Nate Taylor, a graduate student, spent hours making a 3-D computer model of the “peanut,” the mock-up of the nuclear device that North Korea detonated in September. But, because the device was photographed on an angle, not straight on, he had to get creative to take an accurate measurement. He figured out the size of Kim Jong Un’s face by comparing it to basketball player Dennis Rodman’s, and used those numbers to calculate the length and diameter of the bomb. (Anna Fifield/The Washington Post)