Recently, another group of people milling around in a satellite photo attracted international headlines. The U.S.-Korea Institute at Johns Hopkins analyzes satellite photographs of nuclear test sites in North Korea. Earlier this week, they discovered an unusual bit of activity: It appeared that North Korean soldiers stationed at the nuclear test facility were playing volleyball. Three games, in fact, all at the same time. The researchers assessed the implications of this, including that North Korea may have been intentionally trying to signal that the facility was in a stand-down mode. After all, the North Koreans know when commercial satellites are passing overhead, and they may have chosen that moment to send a statement.
As interesting as that is, there’s another component to the story that’s worth considering. The Johns Hopkins team deduced that the North Korean soldiers were playing volleyball mostly given how they were spread out on flat, open areas. In the images, each individual soldier occupies about three pixels square, a tiny space that makes identifying people (much less sports) tricky. You can see a similar issue in that Google Earth image. A person there is about 3 pixels, too. Here, we’ve zoomed in a bit.
It raises the question: Why don’t we have better, more clear imagery that can allow us to see precisely what those North Korean soldiers were doing? Recognizing the complexity of putting a satellite into orbit and using it to photograph the ground, why can’t we at least get a clearer photograph?
The answer? We can — at least if by “we” you mean the government.
I spoke by phone with Raechel Bianchetti, an assistant professor of geography at Michigan State University whose scientific expertise is in satellite image analysis. She explained the constraints that exist in satellite imagery and the ways in which such images have evolved over time.
One of the earliest uses of aerial imagery in a geopolitical context was in World War I. (The Union used hot-air balloons during the Civil War to track Confederate movement but didn’t take photographs for obvious reasons.) Bianchetti pointed me to a collection of aerial photos taken during that war that show military activity at an only slightly worse scale than the images we have of North Korea. Below, an image from the Battle of Cantigny, with the military’s interpretation of the image overlaid.
Those images were taken by the U.S. Air Service, Photographic Section. During the war, surveillance was conducted by biplane, balloon — and even by kite. These days, we tend to forgo lifting people into the air using kites, but we do still rely on aerial photographs.
The company Vexcel Imaging, a subsidiary of Microsoft, makes airplane-mounted cameras that can be used to capture images and other information — like distance data — from midflight. Bianchetti noted that Microsoft’s Bing mapping system used planes from a firm called Keystone Aerial Surveys, which were equipped with Vexcel Ultracams for imaging.
Publicly available images from plane-mounted cameras offer more detail than satellite-mounted ones. DigitalGlobe, the leader in satellite imagery, offers public images that display detail at 30 centimeters-per-pixel. The Vexcel images are clearer than that.
Thirty centimeters per pixel, incidentally, is 90 centimeters for three pixels — or about three feet. This is why those North Korean soldiers were about that size in the Johns Hopkins images. The satellite photos that were being studied were provided by DigitalGlobe.
So why not use higher-resolution images? One reason is that hiring a plane to fly over North Korea isn’t particularly feasible. During the Cold War, the government used Lockheed U-2 spy planes to surveil the Soviet Union, yielding images with a higher resolution than 30 centimeters-per-pixel. There was another drawback made clear when a U-2 piloted by Francis Gary Powers was shot down in 1960. The government could use drones; unmanned aerial vehicles (UAVs) are used in imaging work in the private sector, of course. But why do that if you can get good resolution from space?
After all, the resolution that DigitalGlobe uses is set at 30 centimeters by law, not by technology. In 2014, the federal government cleared DigitalGlobe to sell photos at a higher resolution than the 50 centimeters-per-pixel it had previously been limited to. Its existing satellites — including WorldView-4, which was launched in November, and WorldView-3, launched shortly after the government cleared the new resolution — collect 30-centimeter images that it then sells to the public. For those interested in peeking into North Korea’s back yard, that’s as good as it gets.
Unless you have security clearance.
“I’m fairly certain that, within the government realm, the National Geospatial-intelligence Agency … probably has much higher resolution available to themselves,” Bianchetti said. The resolution that can be achieved is linked to the physical dimensions of the mirror inside the satellite, which has a number of other implications, given that size and weight matter when deploying objects into orbit.
A request for more information from the NGA was not returned by the time this article was published, but it’s safe to say that the precision of U.S. spy satellites is a closely held secret. Details of the once-secret KH-11 satellite design became public when the National Reconnaissance Office donated two of its old devices to NASA in 2012. One estimate of the resolution of the successor to the KH-11 puts the possible resolution in the range of 10 centimeters-per-pixel. In our phone conversation, Bianchetti described having read about a satellite capable of achieving a 7 centimeter-per-pixel resolution.
In short, then, the available resolution for photographs of North Korean soldiers at play is a function not of technology but of security. We will, for the time being, have to be content with being able to see a recent view of nearly any place in the world from anywhere we happen to be, instead of being able to additionally then see what those tourists at the White House happened to be wearing.