These Satellite Images Document an Atrocity

By Robin Mejia
Sunday, June 10, 2007

Before December 2005, the Chadian village of Bir Kedouas was a tidy collection of huts in walled compounds and cultivated fields. A later satellite image shows what is left: The former homesites (marked with red circles) and fields are now a charred scar in the earth. The entire population was either killed or fled. Such images may be providing a powerful new weapon in the struggle to stop genocide.

ON THE CORNEROF JEREMY NELSON'S L-SHAPED DESK AT AMNESTY INTERNATIONAL'S WASHINGTON OFFICE SIT TWO 17-INCH MONITORS. Both have an extra-tall base, and even then they rest on phone books to get them close to eye level for the 6-foot-3 researcher. At 7 on a Thursday night in April, an exhausted but upbeat Nelson is staring at two satellite images of the same area in South Darfur, Sudan, one on each screen. One was taken in December 2004 and the other in February 2007. They show a region that was targeted by what Sudan's government called a "road-clearing offensive." Amnesty reports indicate that last November, while officials were engaged in peace talks in Nigeria, military ground and air forces and Janjaweed militias burned dozens of villages.

This is just one incident in a violent conflict that has killed between 200,000 and 450,000 people in the Darfur region since 2003. A major obstacle to international intervention has been the Sudanese government's refusal to acknowledge the level of violence and its own complicity. In late March, for example, Sudanese president Omar Hassan al-Bashir said in a TV interview that the U.S. State Department map showing 1,000 Darfurian villages as burned was a fabrication. The Sudanese government also holds that only 9,000 people have died in the bloodshed, and that local Janjaweed militias -- the same Janjaweed that gained notoriety for atrocities in Sudan's recent civil war -- are independent actors, despite the fact that they've been seen attacking with military support, raping women and girls, pillaging and sometimes burning entire villages to the ground. Amnesty had been documenting the violence, but last year, the government stopped letting Amnesty's researchers into the country. Then last month, Sudan cited lack of evidence in refusing to comply with the International Criminal Court's arrest warrants for the minister of state for humanitarian affairs and a Janjaweed militia leader.

Which is where Nelson comes in. The 31-year-old researcher is an associate with Amnesty's Crisis Prevention and Response Center. Normally, he tracks hot spots and brewing crises, handles logistics and develops graphics for postcard and poster campaigns. Now he's also learning to analyze satellite images. The catalyst is a partnership between Amnesty and the American Association for the Advancement of Science (AAAS) that's pioneering a new kind of human rights observation: the use of high-resolution satellite imagery -- commercially available only since 2001 -- to document atrocities in areas made inaccessible to watchdog groups. The unusual collaboration started about a year ago, with test projects looking at Zimbabwe and Lebanon. The Darfur effort is by far their biggest yet, and the most politically significant.

Since 2004, the African Union has maintained a modest force of about 7,000 peacekeepers in Darfur, but their mandate expires at the end of this month. The AAAS/Amnesty group hopes that the satellite images it is collecting will provide incontrovertible proof of burning and destruction. Ideally, Sudan then will be forced to accept the United Nations peacekeeping force that the government refused to allow in last year.

"What this satellite technology does, it makes it possible to break down those walls of secrecy. Not only to get information, but to get information in a way that's irrefutable," says Larry Cox, executive director of Amnesty International USA.

The images on Nelson's screen were taken by QuickBird, a satellite launched in 2001 by Colorado-based DigitalGlobe, one of two U.S.-based commercial satellite companies. QuickBird's resolution is good enough to show individual houses and sometimes even cars, and it shoots in color. The other U.S. company, Dulles-based GeoEye, operates two similar satellites with slightly lower resolution. Each of the satellites orbits the planet several times a day; among the three, they reach almost any spot on Earth about once a week. It's fair to assume that government spy satellites still have the best equipment in orbit. But today, anyone with a big enough checkbook can order spy-quality images. (With some exceptions; a 1997 U.S. law prohibits the collection and release of satellite imagery of Israel with a resolution better than two meters, for example.)

Tonight, Nelson begins his work by making a copy of the shot in the right-hand screen and pasting it directly over the one on the left. Then he makes the top one nearly transparent. A river that cuts through the scene becomes a marker to help him line up the two. Now he can easily flip back and forth to look for changes.

Sudanese huts tend to follow a similar pattern: a solid base ring with a steep, thatched roof. In the earlier image, they show up as small circles, with a slight shading to the dome, depending on the direction of the sun. Nelson draws a small, green circle slightly larger than the area of the average hut and makes several dozen copies of it. Then he begins methodically placing a green circle over every hut that can be found in any of the half-dozen settlements spread across the desert landscape.

When he finishes, he moves the 2007 shot to the top and begins the analysis again. When the roof of a thatched hut burns, the base often survives, leaving a telltale ring. But parts of this region were burned so thoroughly that there's nothing left but a large black scar. If you didn't know that huts were there before, you'd have no idea they were now gone.

"Whoever did this did a good job," he says quietly. "Thorough, at least."

By 8 p.m., he has a final tally: Out of 461 structures in the "before" image, most of them homes, 339 were destroyed, five were probably destroyed and 117 were intact. He couldn't tell whether those 117 were still inhabited.

THE ORIGINS OF THE AMNESTY/AAAS COLLABORATION DATE TO OCTOBER 2004, when Nelson's boss, Ariela Blätter, the director of Amnesty's Crisis Prevention and Response Center, got an invitation to a panel discussion on the crisis in Darfur. AAAS had invited speakers from the U.S. Agency for International Development and the State Department to discuss how they'd used high-resolution satellite imagery to map refugee flows. A few years before, the Amnesty group in Denmark had tried using satellite imagery to analyze fire patterns in Darfur, but the imagery had proved too ambiguous to have much of an impact. The State Department's shots, it turned out, were much clearer -- and they came from a commercial satellite, presumably one anyone with funding could access.

"Is this something the human rights community could use?" Blätter remembers raising her hand to ask.

Lars Bromley, a tech-savvy AAAS geographer, was also in the audience. The then-29-year-old understood exactly what the State Department had done to create its report and what it would take for him and AAAS to do the same thing. All he needed was detailed knowledge of where problems were unfolding -- the kind of information researchers at Amnesty could provide.

"He dragged me off of the conference for this intense conversation," recalls Blätter, now 33. "It quickly became clear that they had been looking for me, and I had been looking for them."

Hoping to appear serious about technology herself, Blätter asked him if he'd seen Amnesty Denmark's study. "He told me it was cute," she says.

With Blätter's help, Bromley got funding from the MacArthur Foundation for a pilot project. Darfur was on both their minds, but it was too complicated a situation to try as a test case. So they decided to look at Zimbabwe. In the summer of 2005, President Robert Mugabe's government forces had razed settlements around the country, leaving thousands homeless. The areas targeted had been those that had voted heavily for the opposition, opposition leaders said. Amnesty and the United Nations issued a small mountain of reports describing the nature and scale of the destruction, and documenting the government's subsequent denial of access to aid organizations. Still, the Zimbabwean government insisted that the operation was an urban renewal project with no political agenda.

Blätter and Bromley decided to get satellite images of the settlements before and after the razing. There were plenty of archival shots available for purchase and, through his MacArthur grant, Bromley had money to commission new acquisitions. The only obstacle was getting exact locations: Amnesty has traditionally focused on the personal stories of eyewitnesses, not latitudes and longitudes. So the duo had to be resourceful. Amnesty's London researchers were able to map a settlement called Porta Farm by scanning Google Earth for a site fitting its description: on the main road out of Harare, going toward Bulawayo, between two lakes. For harder-to-find locations, Bromley created a map of the area and e-mailed it to local activists, who e-mailed coordinates back. Bromley placed the order and crossed his fingers.

"You sit there and wonder, 'Did I just waste X amount of dollars on images of the beautiful Zimbabwe countryside?'" he says.

But the shots were spot on. Porta Farm, for example, had consisted of more than 850 buildings that had housed at least 6,000 people. Now, except for dim traces of old dirt tracks, it was an empty landscape. In May 2006, Amnesty issued a news release and, with AAAS, put the images online. Their power quickly became clear. Even though the attacks had happened the previous year and were well-documented, this publicity push generated more coverage for the situation in Zimbabwe -- from outlets including the BBC and al-Jazeera -- than Amnesty had in the previous 10 years. Amnesty's Zimbabwe campaign staff started giving interviews at 1 a.m., coinciding with the news release, and kept going till 10 that evening.

Then that summer, a group called Zimbabwe Lawyers for Human Rights, which through Zimbabwe's court system had been unsuccessfully fighting the forced evictions, submitted the images as evidence in a complaint filed before the African Commission on Human and Peoples' Rights. Mugabe's government seems to have been caught off-guard; officials requested a delay to have the images independently analyzed.

In the aftermath, both Amnesty and AAAS agreed enthusiastically to support the continued partnership, and the MacArthur Foundation funded an expansion. This grant covered Bromley's salary for the next three years. (Researchers at AAAS, like those at most scientific institutions, are expected to cover part or all of their salaries through research grants.) It also allowed Bromley to bring on interns to focus on new areas, including Darfur. A smaller grant, from the Open Society Institute, would pay for a look at Burma. Blätter got a grant from the Save Darfur Coalition, and committed Jeremy Nelson to the project. She also started thinking about using the technology to explore other regions that Amnesty researchers couldn't physically reach. In Eritrea, for example, satellite images might help to locate makeshift jails deep in the interior, where she has heard that political prisoners are held in secret.

ON FRIDAY, BROMLEY, NELSON AND BLÄTTER ARE ON A CONFERENCE CALL. The Amnesty International and AAAS offices are only a few minutes from each other, but this is how most of the group's work gets done.

This is the next phase of the project: They're creating an endangered list of sorts, villages Amnesty thinks are threatened in Darfur and across the border in Chad, where the conflict has spread. The hope is that satellite images of these towns will generate enough media attention to provide some protection.

"We'll let the Sudanese government know that we're watching," Blätter notes. "And asking the global community to join us."

Over the past week, London-based Julie Flint has been gathering intelligence. A researcher and independent journalist who has been working on, and sometimes in, Darfur since 1992, Flint has provided most of the location information the researchers have used for this project. Flint communicates with the Washington team mostly by e-mail, and she tends to send each thought as she has it. Nelson, Blätter and Bromley are used to waking up to an inbox full of one- or two-sentence missives -- locations where Janjaweed militias have been seen massing, updates on which towns one side or the other appears to be taking an interest in, word of where attacks are likely to occur soon. For today's call, Nelson has consolidated the bits on potential attacks into one document, giving the names and locations of towns Flint thinks are at risk.

Before they begin discussing the towns, Bromley notes that each image of Sudan might cost less than he'd expected. Bromley had ordered all of the "before" and many of the "after" images of attacked villages out of DigitalGlobe's archive catalogue. But when he inquired about getting new imagery, he found out that QuickBird was booked solid over Sudan until well into summer.

Apparently, someone with deep pockets is very interested in Sudan, though whether it's government or private enterprise is impossible to know -- DigitalGlobe doesn't release customer information.

There's an upside, however: Users don't purchase the satellite images they requisition; they license them. So after an image is delivered to the original buyer, it'll go into an archive. This system has provided the historical coverage that has allowed Amnesty and AAAS to collect "before" shots.

Because someone else is directing QuickBird, Bromley had to choose another satellite, one of GeoEye's, for the new shots of Darfur. Then, suddenly, that satellite died; apparently, a small component between the satellite's sensor and its memory failed. So Bromley located a third satellite, one run out of the Netherlands Antilles by a company called ImageSat. It can capture almost the same level of detail as QuickBird, but its cameras record in black and white. Blätter and Nelson are worried that these images may not resonate with viewers as deeply as the color shots, but they're the best the team can do for now.

That's when the possibility of a price reduction appeared. From a message Bromley just received, these images of Sudan may each cost about $1,600 each, $900 less than expected. (Satellite companies don't post their rates, but the group has been paying about $2,500 a shot.) Odd -- but good -- news, because at the end of the day, this project has a budget. Blätter has $50,000, about enough for 20 shots, to spend on the threatened villages. And as the point is to let Sudan's government know that Amnesty can order more anytime, she can't use up the whole budget immediately. With this lower price, she'll be able to get more, but there's just not enough money for every village on the initial list.

"So we have to make some hard decisions," she says, beginning the discussion. Bromley's looking at his list as she talks. Like Blätter and Nelson, he hasn't slept much this week. A half-eaten Pop-Tart is pushed off to the side of his desk, and two Starbucks cups sit near his monitors.

"Kafod -- no one has gotten in since 2006," Blätter notes. In theory, at least, even oppressive governments don't want their people to starve; they often let humanitarian aid groups such as Oxfam operate where they don't welcome Amnesty or Human Rights Watch. But aid agencies are occasionally kicked out or, more often, forced to evacuate because of threats to their workers' safety. (After Doctors Without Borders released a report documenting rapes in a Darfur refugee camp in 2005, Sudan issued an arrest warrant for the group's country chief.) Aid agencies haven't been able to enter Kafod for about a year.

"I like it," Bromley says, "If we can get a current image, it can be used by humanitarian organizations."

Bromley writes "yes" next to Kafod, and the discussion moves on. Abu Sakim is a smallish settlement in North Darfur, not far from Kafod. It's under control of the Sudan Liberation Army, one of the main rebel groups, but it's just a few miles south of a Janjaweed stronghold, and militias are believed to have designs on it. Blätter points out that many people believe -- mistakenly -- that it's quiet in northern Sudan; a focus on Kafod and Abu Sakim might help change that perception. As a bonus, it looks as though the two towns might be close enough to each other to capture in a single shot.

Then there's Boldong, a town on the side of a fertile mountain that was attacked and burned early in the conflict but has since been rebuilt. A center for state-run logging operations before the war, Boldong is a strategic site. It's now held by the rebels, but according to Amnesty's information, the government wants to retake it, possibly to restart logging operations.

The team definitely wants Boldong.

Next up is Fanga. Near the same fertile soils as Boldong, the town has been attacked several times, but so far, it looks as though the Janjaweed and the government forces still haven't occupied it. Bromley notes that because of the multiple attacks, new destruction might not be as clear. Blätter says she's going to put parentheses around it.

Blätter sighs. "I want to get all of them," she says. But that's not an option. She pauses and then, in a businesslike tone, says she'll think some more before making a decision on Fanga.

And so it goes, down the list. Bromley stops at Silea. Flint had described a town of at least 500 households, but Google Earth shows only a few scatted huts. He asks Nelson to double-check the location.

SATURDAY AFTERNOON, BROMLEY IS BACK AT HIS NEW YORK AVENUE OFFICE. The night before, he discovered a typo in the list they'd drawn up -- one that could have meant putting in a $2,500 order for the absolutely wrong location. This has been a rushed week. So today, he's going to redo the list from scratch, quietly and free of distractions, and make sure he gets the same coordinates.

Before he maps, he pulls up a program he calls his "fuzzy matcher." It's based on a set of village names and coordinates for Sudan generated by the United Nations. When he types in a name, the program searches for exact or, as the name suggests, fuzzy, matches, returning all possibilities and their official coordinate information. He types in "Boldong."

"Now this is what you want to see," he says. "One 'Boldong' and where we expect to find it."

The phone rings. It's Nelson calling from across town. Over the past two days, both men repeatedly measured the distance between Abu Sakim and Kafod. It turns out the towns are about 6 miles apart. If the researchers put in a special, slightly more expensive order, they might get both in one shot. But there's also a chance the satellite would capture just one town at the very far edge of the frame, or, even worse, that it would only get half of one town. The risk's just not worth it. Abu Sakim's out.

"That's tough math right there," Bromley says. "Hopefully, the information makes your decision for you. But sometimes you have to choose one over the other, and literally you're just looking at a Word document." (Days later, the team will learn how good a decision this was: Abu Sakim had made the list by mistake -- by the time they were discussing it, it had already been overrun.)

Moving down his list, Bromley types "Hashaba" into the fuzzy matcher. On this one, there are 36 near-matches.

By late afternoon, Bromley has what looks like a solid coordinate list for all of the threatened villages, so he turns his attention to the images he has in hand. He's going to convert them into Google Earth files, viewable by the 100 million people or so who have downloaded the program.

He pulls one up. It could be an archaeological site: Small, round outlines mark where buildings once stood; old fence lines are still slightly visible as marks in the sand. With no context, the scene is beautiful. But with a little bit of background, it's devastating. Those small, round outlines recently were homes. The people who lived there, if still alive, are probably in refugee camps. You can see those in other images: tightly packed huts and tents that house tens of thousands of people in squalid conditions. In one image taken in February near a village called Tawila, the huts are built right up to the fence line of an African Union outpost, apparently in the hope of garnering a bit of extra protection.

BROMLEY STAYS HOME ON SUNDAY. He can log into his work computers from there; he needs a good night's sleep and some time with his dog. Nelson isn't so lucky. He hasn't taken a full day off in more than two weeks, but his work computer isn't accessible from off-site, and he has another half-dozen image pairs to analyze, so he's at the office again, though dressed down in jeans and a Nationals cap.

At midnight, Blätter calls to check in. "Oh, that's not good," she says when Nelson picks up. She's worried about the hours he's keeping, and was hoping she'd reach his voice mail. He promises to call it a night.

By Monday afternoon, the team has accomplished a lot. Blätter has made the final decisions about the threatened sites and has reviewed a thick stack of documentation on the villages that have already been attacked. Julie Flint has just returned from Chad, where she had interviewed refugees, collecting heartbreaking photos and testimonials that will humanize the satellite shots when Amnesty posts them online. Nelson's reviewing the information that will be shipped to the graphic design firm responsible for the group's Web site.

With his part of the project under control, Bromley takes a break from Darfur. Just down the hall from his office is the AAAS Burma conflict-monitoring center. Intern Sean O'Connor, a 25-year-old who's starting grad school in the fall, has a desk that is separated from the hall by a tall counter. Bromley leans on this when he stops by to discuss a package they're getting ready to send to their contacts in Thailand.

O'Connor has been working with Bromley to document the Burmese government's persecution of the Karen, an ethnic minority group that lives along the country's mountainous border with Thailand. If anything, this task has proved even more challenging than documenting the destruction in Darfur: Rather than huts in a desert, the targets are homes in a jungle, in a part of the world often hidden by clouds. But the duo has documented a number of attacks and, a few weeks ago, posted early findings on Google Earth. The activists they'd been working with were impressed but afraid that the detailed information the images provided might help the military find local hideouts. So Bromley took the files down, and now they're mailing them on a DVD, along with a paper printout of a satellite image that could be taken to a refugee camp.

"This is called participatory mapping," Bromley says. "Rather than me speculating what things are, they can say, 'Look, there's my house.'"

As the week moves on, O'Connor gets information from the Free Burma Rangers that a one-day government offensive has just burned down four Burmese villages, leaving about 1,000 people homeless. The towns are small and close to each other, so he puts in a satellite order that should encompass all four. But the formal start of the monsoon season is less than two weeks away, and cloud cover is already becoming a problem. QuickBird may not get a clear image until next week, next month -- or even in the fall, after the monsoons. By fall, any burn scars will have been overgrown by jungle. Two of the villages aren't in DigitalGlobe's library; unless new images are taken soon, there won't be evidence that they ever existed.

As O'Connor tracks the situation in Burma, the images for the at-risk Darfurian villages start to come in. It turns out that Bromley was mistaken about some images costing less than expected. But after Bromley explains the project, the company offers him a deal: For every 10 new images ordered, two additional ones will be free. The other companies Bromley works with have given him great deals on their archival images -- one has even donated some -- but this is the first break on new collections. It means the group will be able to afford to shoot most, though not all, of the threatened sites on their list.

As the first set of images from the Netherlands satellite arrives, Bromley encounters another glitch: Some aren't crisp enough. Satellite companies list the resolution their cameras get when they're pointing straight down at the ground, but satellites often end up shooting at a slight angle for speed. The larger the angle, the lower the resolution. Depending on why one wants an image, the difference may not matter. But Darfurian huts are small.

"If it's too fuzzy, we can't see if they're damaged," Bromley says.

Fortunately, the company agrees to redo the shots. The next week, those images arrive, and they're what he'd hoped for: crisp, unassailable evidence of villages that still exist -- and proof to Sudan's government that someone is watching.

NEXT: See the satellite images.

Robin Mejia writes about science, technology and people. She lives in Santa Cruz, Calif., and can be reached at

View all comments that have been posted about this article.

© 2007 The Washington Post Company