This satellite image taken Tuesday, June 20, 2017, and released by the National Oceanic and Atmospheric Administration shows a tropical storm over the Gulf of Mexico approaching the Gulf Coast.

This satellite image taken Tuesday, June 20, 2017, and released by the National Oceanic and Atmospheric Administration shows a tropical storm over the Gulf of Mexico approaching the Gulf Coast. AP / NOAA

Detecting Secret Military Exercises With Micro Satellites, a How-To

The future of intelligence is small teams and tiny satellites. It’s not a future the U.S. will own exclusively.

Detecting an adversary's unscheduled naval exercise used to require a massive operation perhaps involving multiple agencies. In March 2017, a small team of analysts with a firm called 3GIMBALS, working with U.S. Southern Command, used imagery from a constellation of inexpensive microsatellites to pick up an unscheduled military exercise in Venezuela that had taken place a year prior. How they did it says a lot about the future of intelligence collection from space, where the U.S. has a rapidly shrinking advantage.

The game-changer wasn’t the algorithm; it was the availability of cheap, daily satellite imagery — in this case, from a startup called Planet Labs. [Full disclosure: Defense One has an agreement to get limited free imagery from Planet.] With a constellation of 149 microsatellites, Planet can cover a lot more territory than any single high-end satellite, so almost anywhere in the world you look, Planet can find a recent picture of it. With new imagery coming in every day, you can automate a search for changes at various locations. That means you can detect when a military begins to operate in a given area, and perhaps even deduce what kind of equipment they’re using — all at exponentially less cost than it would take to have run the same operation a decade ago.

In 3GIMBALS’ case, analysts trained an algorithm to track changes in the quantity and position of white and blue pixels in a given area. That might not sound particularly revolutionary, but it enabled them to verify that the number of military vessels was unusually low during a window of time in March 2016. This meant vessels were missing from a strategically important port. Scouring for additional open-source clues, 3GIMBALS verified that Venezuela was running an unscheduled and unplanned military exercise — a response, in part, to President Obama’s March 4 re-declaration of Venezuela as a national security threat. They were even able to deduce what types of boats were engaged in the exercise and whether they were new.

“We’re able to validate and detect a boat that was being manufactured in Vietnam and shipped to Venezuela. We can then confirm or deny through Planet imagery what Venezuela publicizes on YouTube about delivery and production schedule of that vessel,” said Gerby Marks, an analyst with 3GIMBALS. “Planet’s global ‘always on’ collect can help us detect and validate movement of that vessel from that location.”

It’s not the first time a small group of analysts have used open-source satellite data to reveal, or even predict, a military operation. In 2011, a group of students with Harvard’s Satellite Sentinel Project, or SSP, predicted that the Sudanese Armed Forces were about to invade the disputed area of Abyei. The tip-off was the construction of new, broad roads – of the sort that might be used to carry oil equipment, or military units – in  a place where there was no oil. “These roads indicated the intent to deploy armored units and other heavy vehicles south towards Abyei during the rainy season,” the group wrote in their report. The Sudanese invasion began a few months later.

But that project got some special help in the form of George Clooney, who was laser-focused on Sudan. The ability to run similar analysis virtually anywhere at anytime, and then train simple algorithms to identify small daily changes in the images turns what used to be a difficult and expensive intelligence activity into something that a small team can perform. Planet Lab’s Dove satellites, each about the mass of a pair of toasters, are a fraction of the size of conventional surveillance satellites.

“This always-on imagery, and the automation of monitoring and knowing this was a significant blip right in the norm. We were able to go and say, what’s happening there. That's the gist of the game-changing aspect of Planet. It’s always there, we don’t need to know where we want to go look,” said Maj. Mike Little Sr., who is part of U.S. Southern Command’s Joint Intelligence Operation Center-South,  GEOINT Division. “We didn’t know to look. When we ran an algorithm that showed a big dip in the military vessels, nobody knew that was going to happen. That was in response to President Obama’s declaration about Venezuela. They did not even know to go look and watch that Venezuela was going to go do something.”

Military leaders say that small satellites, defined as weighing less than 500 kilograms and so far limited to relatively low-resolution imagery from low Earth orbit, aren’t going to replace high-resolution satellites like DigitalGlobe’s Worldview 4.

“Big sats are still very important to us. Our exquisite systems, there’s no replacing them,” Col. Steve Butow, the military lead for the Defense Innovation Unit Experimental, or DIUx, said at a CSIS event on Wednesday. Yet Butow also expressed optimism about how new, small satellites for intelligence would enable the military to much better monitor emerging crises. He called global threat monitoring “the ultimate big data problem. We need data from everywhere, not just particular targets.”

DIUx is working with a different small-satellite company, Capella Space, which uses synthetic aperture radar to collect data all over the world. The company, based in Silicon Valley, will launch its first satellite next year.

Lots and lots of low-cost cameras and other sensors floating miles above the Earth could be a huge asset. But it won’t be unique to the United States.

The number of small satellites is expected to surpass 1,500 by the end of the year, according to recent numbers from Avascent Analytics. “There are 70 countries that are one way or another involved in small sats,” Bhavya Lal, a researcher at the Science and Technology Policy Institute at the Institute for Defense said at Wednesday’s CSIS event. “We need to be prepared for the possibility these companies will be going overseas” where they will be customized for overseas customers, she said.

Southcom’s Little put it a bit more bluntly. “Our competitors will learn to use this exploding geo-enabled capability to learn a lot of things. If we don’t use this as well as well as them and use our exquisite assets to be a differentiator and stay ahead, we’re going to get behind.”